A global multi-source composite verification hydropower plant sub-scene hierarchical security linkage system and method

Abstract

The application discloses a global multi-source composite verification hydropower plant sub-scene hierarchical security linkage system and method, belonging to the technical field of intelligent hydropower plant security emergency, comprising: a global multi-source heterogeneous sensing collection module accesses hydropower plant heterogeneous equipment through a protocol gateway and collects multiple types of sensing data; a multi-source composite alarm verification module performs video AI and sensor timing joint verification for preset multiple types of risk scenes, multi-source synchronous triggering outputs an alarm containing a scene, coordinates and confidence, and single-source triggering filters out interference; a hydropower sub-scene hierarchical linkage strategy scheduling module matches a three-level strategy library and issues corresponding instructions; a BIM dynamic evacuation and onCall hierarchical pushing module dynamically generates an evacuation route based on BIM and UWB positioning and hierarchically pushes an alarm in multiple channels; and an alarm closed-loop trusted notarization self-optimization module performs distributed notarization on global link data hash, generates a disposal work order archive and iteratively updates verification parameters in a Bayesian model.

Description

Technical Field

[0001] This application relates to the field of security monitoring technology for intelligent hydropower plants, and in particular to a multi-source composite verification system and method for security linkage in hydropower plants across different scenarios and levels. Background Technology

[0002] Existing alarm schemes based on single sensors have a high false alarm rate in the actual operating environment of hydropower plants: the long-term high humidity environment in underground powerhouses makes smoke detectors susceptible to water vapor interference; industrial television has poor stability in extracting flame and smoke features under strong light reflection and airflow disturbance conditions; and a single source of information is difficult to reliably distinguish between real accidents and environmental interference in the complex environment of hydropower plants.

[0003] In the existing fragmented security system with multiple subsystems deployed independently, video surveillance, environmental sensing, access control and fire monitoring systems are built independently by different suppliers. The data of each system is not interconnected, the clocks are not synchronized, and the linkage logic is fixed. It can only realize open-loop control of "sensor triggering → single device action" and cannot perform differentiated linkage according to the type and severity of the accident.

[0004] For non-fire scenarios such as SF6 leaks and reservoir intrusions, there is generally no corresponding linkage logic coverage. Existing solutions lack real-time personnel location and dynamic evacuation planning capabilities, and emergency evacuation relies on pre-drawn static escape maps, making it difficult to adjust routes in real time according to actual risk points, which is particularly dangerous in narrow underground corridors. In terms of event tracing, existing solutions generally use local database logging, and alarm records and execution logs can be manually modified or deleted, failing to meet the compliance requirements for data integrity in power industry safety audits; at the same time, the adjustment of false alarm thresholds in existing systems relies entirely on manual on-site debugging, lacking the ability for independent iterative optimization.

[0005] Therefore, there is a need for a security linkage system and method that can achieve multi-source perception composite verification, scenario-based hierarchical differentiated linkage, dynamic evacuation based on real-time positioning, full-process reliable evidence storage, and adaptive optimization for various risk scenarios across the entire hydropower plant area. Summary of the Invention

[0006] The purpose of this invention is to provide a multi-source composite verification system and method for security linkage in hydropower plants, categorized by scenario and level, to solve the aforementioned technical problems in the prior art.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: The present invention provides a multi-source composite verification and scenario-based security linkage system for hydropower plants, comprising:

[0008] The multi-source heterogeneous sensing and acquisition module connects to heterogeneous sensing devices deployed in underground powerhouses, substations, reservoirs, and cable corridors of hydropower plants through multi-industry protocol gateways. It collects video, radar, gas concentration, liquid level, and UWB positioning data, and connects to the maintenance work order system, personnel access control system, and BIM database to output the data after unified parsing of the multi-source data.

[0009] The multi-source composite alarm verification module is connected to the global multi-source heterogeneous sensing and acquisition module. For six types of risk scenarios, namely fire, SF6 leakage, flooded factory, high voltage overpass of substation, reservoir intrusion and construction personnel overpass, it performs video AI feature recognition and time-series joint verification of corresponding category dedicated sensors. When multiple source signals are triggered synchronously within a preset time window, it outputs an alarm signal containing the risk scenario, risk coordinates and confidence score. When a single source is triggered, it is determined to be environmental interference and filtered out.

[0010] The hydropower scenario-based hierarchical linkage strategy scheduling module is connected to the multi-source composite alarm verification module. It has a three-level linkage strategy library pre-stored in the module, which corresponds to the six types of risk scenarios. Based on the received alarm signals, it matches the corresponding scenario and level, generates ventilation, access control, power supply, fire protection and gate control commands, and sends them to the on-site execution equipment.

[0011] The BIM dynamic evacuation and onCall hierarchical push module is connected to the multi-source composite alarm verification module. It includes a BIM dynamic evacuation submodule and an onCall hierarchical push submodule. The BIM dynamic evacuation submodule accesses the factory's 3D BIM drawings and UWB real-time personnel positioning data. When a preset alarm level is triggered, it avoids the risk area with the risk coordinates as the center and generates an evacuation route based on the corridor topology. When the risk range expands, it switches to a backup channel. The onCall hierarchical push submodule matches the corresponding push level according to the alarm level and distributes alarm information synchronously through multiple channels.

[0012] The alarm closed-loop trusted evidence storage and self-optimization module is data-connected to the hydropower scenario-based hierarchical linkage strategy scheduling module. It includes a hash evidence storage unit, a work order closed-loop unit, and a strategy self-optimization unit. The hash evidence storage unit generates unique hash values ​​for sensing data, verification results, linkage instructions, execution receipts, and push records, and stores them in a distributed manner. The work order closed-loop unit automatically generates electronic handling work orders in response to alarm signals and receives handling data uploaded by maintenance personnel to complete closed-loop archiving. The strategy self-optimization unit reads historical samples of closed-loop work orders and uses a Bayesian model to iteratively update the composite verification time window parameters and confidence level boundary values ​​for each risk scenario.

[0013] Through a dual verification mechanism combining video AI feature recognition and dedicated sensor timing verification, the system can determine whether multiple source signals are triggered synchronously within a preset time window, thereby effectively filtering single-source environmental interference and outputting a highly reliable alarm signal with a confidence score, providing accurate data support for subsequent hierarchical linkage.

[0014] Furthermore, the preset time window is 3 to 10 seconds, and the time window is dynamically adjusted according to the Bayesian model iteration update results of the strategy self-optimization unit. This time window range takes into account the response delay differences of different types of sensors, which can capture the quasi-synchronous triggering characteristics of multi-source signals in real accidents, and effectively eliminate occasional single-source false triggering caused by equipment jitter or environmental fluctuations.

[0015] Specifically, the time-series joint verification rules for each risk scenario in the multi-source composite alarm verification module are as follows:

[0016] In a fire scenario, an effective alarm is output when at least two of the signals from the industrial TV's flame and smoke AI recognition system are triggered simultaneously with the signals from the smoke detector, heat sensor, and CO concentration sensor.

[0017] In an SF6 leak scenario, an effective alarm is output when the thermal imaging gas cloud identification signal and the SF6 infrared concentration sensor signal are triggered synchronously.

[0018] In the scenario of a flooded factory, a valid alarm is output when the video water accumulation recognition signal is triggered synchronously with the signals from at least two liquid level sensors.

[0019] In the case of high voltage over-limit scenario at the booster station, an effective alarm will be output when the human body AI recognition signal, millimeter-wave radar signal and thermal imaging temperature measurement signal are triggered simultaneously.

[0020] In the scenario of intrusion into the reservoir area, an effective alarm is output when the AI ​​recognition signal of a vessel or person is triggered synchronously with the shoreline radar signal.

[0021] When construction workers cross the boundary, a valid alarm is output when the facial recognition signal, RFID positioning signal, and maintenance work order authorization signal are triggered simultaneously.

[0022] The exclusive verification rules for the above six scenarios are customized for the physical characteristics of the perceived signals in each scenario, ensuring that the verification logic is highly matched with the signal characteristics of actual risks.

[0023] Furthermore, the three-level linkage strategy library is divided according to the confidence score as follows: a confidence score in the first preset interval corresponds to level one risk, a confidence score in the second preset interval corresponds to level two risk, and a confidence score in the third preset interval corresponds to level three risk.

[0024] In response to fire scenarios, Level 1 involves activating local ventilation and issuing a pop-up warning on the central control platform; Level 2 involves adding zoned smoke extraction, unlocking all access control doors in the corresponding fire compartment, forcing elevators to land, and activating emergency broadcasts; Level 3 involves adding automatic power cut-off to the high-voltage power supply to the fire area, activating automatic fire extinguishing devices, setting all underground access control doors and warehouse gates to always open, dynamically switching to backup evacuation routes, and pushing alarms and factory structure data to the higher-level dispatch platform and local emergency platform.

[0025] This tiered linkage strategy enables precise responses to risks of varying severity, avoiding excessive linkage in low-risk situations that could disrupt normal production operations.

[0026] Preferably, the BIM dynamic evacuation submodule, for scenarios involving narrow underground corridors, sequentially unlocks access control points along the corridor's topological path and simultaneously generates evacuation routes in both upstream and downstream directions; for scenarios involving multi-story factory buildings, it generates vertical evacuation routes excluding the floors where the fire originated. The simultaneous generation of bidirectional evacuation routes allows personnel to choose the nearest exit based on their location, significantly shortening evacuation time.

[0027] Furthermore, the onCall hierarchical push submodule has the following push hierarchy: Level 1 alarms are pushed to team maintenance personnel; Level 2 alarms are pushed to team maintenance personnel and safety management personnel; Level 3 alarms are simultaneously pushed to team maintenance personnel, safety management personnel, duty plant manager, superior dispatch center, and local emergency management platform; push attachments include alarm coordinates, real-time video snapshots, BIM evacuation drawings, and monitoring data curves. This hierarchical push mechanism ensures that personnel at all levels receive alarm information matching their responsibilities in a timely manner, avoiding information overload or omissions.

[0028] In some preferred embodiments, the input features of the Bayesian model in the strategy self-optimization unit include the trigger timing difference of each sensor, ambient temperature and humidity, and light intensity. The prior distribution uses historical false alarm rate statistics, and a model update is triggered every preset number of closed-loop work orders. By introducing environmental features as input, the Bayesian model can automatically identify false alarm patterns under different environmental conditions and achieve adaptive iterative optimization of the verification parameters.

[0029] As a preferred embodiment, the system also includes a data interaction interface with the superior dispatch platform and the local emergency response platform. This interface is used to synchronously upload alarm event information, plant BIM structural data, and real-time on-site monitoring data to the superior dispatch platform and the local emergency response platform. This interface enables external emergency command organizations to grasp the plant's risk situation in real time, providing data support for cross-departmental collaborative response.

[0030] Furthermore, the comprehensive multi-source heterogeneous sensing and acquisition module connects to the heterogeneous sensing devices through a multi-industry protocol gateway compatible with Modbus, ONVIF, MQTT, and Power 104 protocols. The acquisition range includes smoke detectors, temperature sensors, CO concentration sensors, SF6 gas sensors, liquid level sensors, industrial video, and UWB personnel positioning data in the underground plant area; millimeter-wave perimeter radar, thermal imaging temperature measurement, and high-voltage switch status data in the booster station area; and shoreline photoelectric cameras, vessel monitoring radar, and water level data in the reservoir area. The use of a multi-protocol compatible gateway allows the system to complete heterogeneous access without modifying existing equipment, reducing system deployment costs.

[0031] This invention also provides a multi-source composite verification method for security linkage in hydropower plants across different scenarios, comprising the following steps:

[0032] S1. Access the heterogeneous sensing devices throughout the hydropower plant via a multi-industry protocol gateway to collect video, radar, gas concentration, liquid level and UWB positioning data, and simultaneously connect to the maintenance work order system, personnel access control system and BIM database to complete unified parsing of multi-source data;

[0033] S2. For six types of risk scenarios, namely fire, SF6 leakage, flooding of factory buildings, high voltage overpass of substation, reservoir intrusion and construction personnel overpass, video AI feature recognition and corresponding dedicated sensor time sequence joint verification are performed within a preset time window. Single-source triggers are judged as environmental interference and filtered out, and multi-source synchronous triggers are judged as valid alarms. The alarm signal includes risk scenario, risk coordinates and confidence score.

[0034] S3. Based on the risk scenario and confidence score of the alarm signal, match the pre-stored three-level linkage strategy library, generate corresponding level ventilation, access control, power supply, fire protection and gate control commands and issue them to the on-site execution equipment;

[0035] S4. When a preset alarm level is triggered, based on the factory's 3D BIM drawings and UWB personnel real-time positioning data, an evacuation route is generated to avoid the risk area with the risk coordinates as the center. When the risk range expands, the backup channel is switched. At the same time, the corresponding push level is matched according to the alarm level, and alarm information including alarm coordinates, real-time video snapshots, BIM evacuation drawings and monitoring data curves are distributed synchronously through multiple channels.

[0036] S5 generates unique hash values ​​for sensing data, verification results, linkage instructions, execution receipts, and push records, and stores them in a distributed manner; it automatically generates electronic handling work orders and completes closed-loop archiving after receiving handling data uploaded by maintenance personnel.

[0037] S6. Read historical samples of closed-loop work orders and use a Bayesian model to iteratively update the composite verification time window parameters and confidence level boundary values ​​for each risk scenario.

[0038] Furthermore, the three-level linkage strategy library mentioned in step S3 is divided into levels according to confidence scores of 30-60 points corresponding to Level 1, 60-85 points corresponding to Level 2, and 85-100 points corresponding to Level 3.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] 1. By constructing a joint verification mechanism for video AI feature recognition and dedicated sensor timing, multi-source signal synchronous triggering is used as an effective alarm judgment condition within a preset time window, and single-source triggering is directly filtered out as environmental interference. This fundamentally eliminates the problem of false alarms from a single sensor in complex environments such as high humidity, strong light reflection, and airflow disturbance in hydropower plants, effectively improving the accuracy of security event identification and overcoming the technical defects of existing single-sensor solutions with high false alarm rates under harsh working conditions.

[0041] 2. By constructing exclusive time-series joint verification rules for six risk scenarios—fire, SF6 leakage, flooding of the plant, high-voltage intrusion at the substation, reservoir intrusion, and construction personnel crossing the boundary—and configuring a three-level differentiated linkage strategy library corresponding to the scenario and confidence score, we have achieved full-scenario coverage and precise graded response for multiple types of risks across the entire hydropower plant, overcoming the technical problems of incomplete scenario coverage and rigid linkage logic in the existing fragmented security system.

[0042] 3. By accessing the factory's 3D BIM drawings and UWB personnel real-time positioning data, when an alarm is triggered, the system dynamically avoids risk areas centered on risk coordinates and generates evacuation routes in real time based on corridor topology. When the risk range expands, it automatically switches to backup channels. This overcomes the technical defects of existing solutions that rely on static escape maps and cannot adjust the path in real time according to the actual risk points, effectively improving the emergency evacuation safety in special scenarios such as underground narrow corridors.

[0043] 4. By matching alarm levels with corresponding push levels and distributing alarm information containing alarm coordinates, real-time video snapshots, BIM evacuation drawings, and monitoring data curves through multiple channels, the system achieves accurate matching of alarm information with responsible personnel at all levels. This avoids the coexistence of information overload and high-risk alarm omissions, and improves the timeliness and accuracy of emergency response.

[0044] 5. By generating unique hash values ​​for sensing data, verification results, linkage instructions, execution receipts, and push records and distributing them, the system achieves tamper-proof evidence storage of data across the entire chain, meets the compliance requirements for data integrity in the power industry's safety audit, and overcomes the technical defects of existing local database logs that can be manually modified or deleted.

[0045] 6. By reading historical samples of closed-loop work orders through the strategy self-optimization unit, and using a Bayesian model to iteratively update the composite verification time window parameters and confidence level boundary values ​​for each risk scenario with the goal of minimizing the false alarm rate, the system achieves autonomous iterative optimization, overcoming the technical problem that the adjustment of the false alarm threshold in the existing system completely relies on manual on-site debugging and has high operation and maintenance costs. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the hydropower plant scenario-based hierarchical security linkage system with full-domain multi-source composite verification in this embodiment.

[0047] Figure 2 This is a schematic diagram of the time-series joint verification process of the multi-source composite alarm verification module for six types of risk scenarios in this embodiment.

[0048] Figure 3 This is a schematic diagram of the execution flow of the hydropower scenario-based hierarchical linkage strategy scheduling module in this embodiment.

[0049] Figure 4 This is a schematic diagram of the structure of the BIM dynamic evacuation and onCall hierarchical push module in this embodiment.

[0050] Figure 5 This is a schematic diagram of the workflow of the alarm closed-loop trusted evidence storage self-optimization module in this embodiment. Detailed Implementation

[0051] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the following embodiments are used to illustrate the technical solutions of the present application, and are not intended to limit the scope of protection of the present application.

[0052] In one implementation, combined with Figure 1 The system architecture shown fully describes the basic implementation process of a multi-source composite verification system for hydropower plants, categorized by scenario and level. This implementation method is applied to a large-scale conventional smart hydropower plant, whose plant area includes underground powerhouses, substations, long and narrow reservoirs, and multiple deeply buried cable corridors. The entire security linkage system of this invention is deployed, and the overall hardware includes industrial video cameras, thermal imaging equipment, various environmental sensors, UWB positioning base stations, access control controllers, ventilation equipment, fire-fighting equipment, reservoir area gates, industrial edge gateways, and a back-end server platform.

[0053] The global multi-source heterogeneous sensing and acquisition module connects to heterogeneous sensing devices deployed in underground plants, substations, warehouses, and cable corridors through a multi-industry protocol gateway compatible with Modbus, ONVIF, MQTT, and the Power 104 protocol. It collects video, radar, gas concentration, liquid level, and UWB positioning data, and interfaces with maintenance work order systems, personnel access control systems, and BIM databases. After completing unified parsing of multi-source data, it outputs the data in real time to the multi-source composite alarm verification module.

[0054] The underground plant area collects data from smoke detectors, temperature sensors, CO concentration sensors, SF6 gas sensors, liquid level sensors, industrial video, and UWB personnel positioning data; the booster station area collects data from millimeter-wave perimeter radar, thermal imaging temperature measurement, and high-voltage switch status data; and the reservoir area collects data from shoreline photoelectric cameras, vessel monitoring radar, and water level data. A multi-industry protocol gateway enables heterogeneous devices to connect without modification, reducing system deployment costs.

[0055] Multi-source composite alarm verification module combined Figure 2 The verification process shown performs time-series joint verification of video AI feature recognition and corresponding dedicated sensors for six risk scenarios: fire, SF6 leakage, flooded factory, high-voltage boundary crossing of substation, reservoir intrusion, and construction personnel boundary crossing.

[0056] In a fire scenario, an alarm is considered valid when at least two of the signals from the industrial TV's flame and smoke AI recognition system and the signals from the smoke, heat, and CO concentration sensors are triggered synchronously within a preset time window.

[0057] In the case of SF6 leakage, when the thermal imaging gas cloud identification signal and the SF6 infrared concentration sensor signal are triggered synchronously, it is determined to be a valid alarm;

[0058] In the scenario of a flooded factory, a valid alarm is determined when the video water accumulation recognition signal and the signals from at least two liquid level sensors are triggered simultaneously; in the scenario of a high-pressure over-limit at a booster station, a valid alarm is determined when the human body AI recognition signal, the millimeter-wave radar signal, and the thermal imaging temperature measurement signal are triggered simultaneously.

[0059] In the scenario of intrusion into the reservoir area, when the AI ​​recognition signal of a vessel or person is triggered synchronously with the shoreline radar signal, it is determined to be a valid alarm;

[0060] In scenarios where construction workers cross boundaries, a valid alarm is determined when facial recognition, RFID positioning, and maintenance work order authorization signals are triggered simultaneously. If only one source triggers, it is considered environmental interference and filtered out, and no alarm signal is generated. A valid alarm output includes the risk scenario, risk coordinates, and a confidence score ranging from 0 to 100.

[0061] The preset time window is preferably 3 to 10 seconds, and can be dynamically adjusted according to the Bayesian model update results of the strategy self-optimization unit.

[0062] Hydropower scenario-based hierarchical linkage strategy scheduling module combined with Figure 3 The execution flow shown matches the corresponding scenario and level based on the received alarm signal. The internally pre-stored three-level linkage strategy library is divided according to confidence score: the first preset interval (preferably 30 to 60 points) corresponds to level 1 risk, the second preset interval (preferably 60 to 85 points) corresponds to level 2 risk, and the third preset interval (preferably 85 to 100 points) corresponds to level 3 risk.

[0063] In response to fire scenarios, Level 1 involves activating local ventilation and issuing a pop-up warning on the central control platform; Level 2 involves adding zoned smoke extraction, unlocking all access control doors in the corresponding fire compartment, forcing elevators to land, and activating emergency broadcasts; Level 3 involves adding automatic power cut-off to the high-voltage power supply to the fire area, activating automatic fire extinguishing devices, setting all underground access control doors and warehouse gates to always open, dynamically switching to backup evacuation routes, and pushing alarms and factory structure data to the higher-level dispatch platform and local emergency platform.

[0064] SF6 leakage, flooding of the plant, high-voltage overtravel at the substation, intrusion into the reservoir area, and construction personnel overtravel are all configured with dedicated three-level differentiated linkage logic to generate control commands for ventilation, access control, power supply, fire protection, and gate access and issue them to the on-site execution equipment.

[0065] BIM dynamic evacuation and onCall hierarchical push module combined Figure 4 The architecture shown includes a BIM dynamic evacuation submodule and an onCall hierarchical push submodule.

[0066] The BIM dynamic evacuation submodule connects to the plant's 3D BIM drawings and UWB real-time personnel location data. When a level 2 or 3 alarm is triggered, it avoids the risk area with the risk coordinates as the center and generates an evacuation route based on the corridor topology. When the risk range expands, it switches to the backup passage.

[0067] For scenarios involving narrow underground corridors, access control doors are unlocked sequentially along the corridor's topological path, simultaneously generating evacuation routes in both upstream and downstream directions, allowing personnel to choose the nearest exit based on their location. For multi-story factory scenarios, vertical evacuation routes excluding the floors where the fire originated are generated. The evacuation routes are simultaneously displayed on the central control screen, on-site terminals, and on the smart safety helmets of the workers.

[0068] The onCall hierarchical push submodule matches the corresponding push level according to the alarm level: Level 1 alarms are pushed to the team maintenance personnel; Level 2 alarms are pushed to the team maintenance personnel and safety supervision personnel; Level 3 alarms are pushed to the team maintenance personnel, safety supervision personnel, duty manager, superior dispatch center and local emergency management platform simultaneously; the push attachments include alarm coordinates, real-time video snapshots, BIM evacuation drawings and monitoring data curves, and are distributed simultaneously through multiple channels such as the central control screen, sound and light alarms, SMS, mobile APP and smart safety helmet intercom terminal.

[0069] Alarm closed-loop trusted evidence storage self-optimization module combined Figure 5 The workflow shown includes a hash-based evidence storage unit, a work order closed-loop unit, and a strategy self-optimization unit.

[0070] The hash-based evidence storage unit generates unique hash values ​​for sensing data, verification results, linkage instructions, execution receipts, and push records, and stores them in a distributed manner to achieve tamper-proof evidence storage of data across the entire chain, meeting the compliance requirements for data integrity in the power industry's security audit.

[0071] The work order closed-loop unit automatically generates electronic handling work orders in response to alarm signals, and completes closed-loop archiving after receiving the handling data uploaded by maintenance personnel.

[0072] The strategy self-optimization unit reads historical samples of closed-loop work orders and uses a Bayesian model to iteratively update the composite verification time window parameters and confidence level boundary values ​​for each risk scenario, with the goal of minimizing the false alarm rate. The input features of the Bayesian model include the triggering time difference of each sensor, ambient temperature and humidity, and light intensity. The prior distribution uses historical false alarm rate statistics. The model is updated once every preset number of closed-loop work orders, so as to realize the system's autonomous iterative optimization.

[0073] In another implementation, taking the emergency response to a level-two fire in an underground plant as an example, the specific execution process of the system is explained.

[0074] The system's industrial video recognition detects smoke and flame characteristics in the computer room. Simultaneously, smoke detectors and CO sensors collect abnormal data. The multi-source composite alarm verification module determines that the multi-source signals are triggered synchronously within a preset time window, outputs a valid alarm, and the confidence score is in the second preset range, thus classifying it as a level 2 risk.

[0075] The system immediately matched the Level 2 fire linkage strategy, and the water and electricity scenario-based hierarchical linkage strategy scheduling module generated and issued control commands: start zoned smoke extraction, unlock all access control doors in the underground plant, force the plant elevators to land, and activate emergency broadcasts. After receiving the alarm coordinates, the BIM dynamic evacuation submodule, combined with the plant-wide personnel UWB positioning data, avoided the fire-prone machine room area and dynamically generated two independent evacuation routes, one upstream and one downstream, which were simultaneously displayed on the central control screen, the on-site broadcast terminal, and the smart safety helmets of the workers.

[0076] The onCall layered push submodule synchronously pushes alarm information to the team's maintenance personnel and safety management personnel, along with video snapshots of the fire location, BIM evacuation drawings, and monitoring data curves.

[0077] The hash evidence storage unit generates a unique hash value for all data related to alarm verification, command issuance, device execution, and message push, and stores it in a distributed manner. The work order closed-loop unit automatically generates a handling work order. After the maintenance personnel complete the on-site handling, they take photos and send them back to complete the closed-loop archiving.

[0078] The strategy self-optimization unit automatically optimizes the flame recognition time window parameters for this area based on the valid alarm samples.

[0079] In another implementation, the system's handling process for high-risk scenarios is illustrated using the example of handling illegal vessel intrusion at the third level of the reservoir area.

[0080] The reservoir shoreline radar detected a moving target on the water surface, and the dual-spectrum AI video recognition confirmed that an illegal vessel was approaching a high-risk area. The dual-source data were simultaneously verified within the preset time window, and the confidence score was in the third preset interval, which determined it to be a level three high-risk reservoir intrusion event.

[0081] The system immediately executes the three-level linkage strategy. The hydropower scenario-based hierarchical linkage strategy scheduling module generates and issues control commands: activates the shore-side audible and visual deterrent equipment, locks relevant operation permissions, and sets the reservoir area gate to always open. The BIM dynamic evacuation submodule marks the intrusion coordinates and high-risk isolation areas in real time and generates evacuation routes for personnel in the reservoir area.

[0082] The onCall layered push submodule synchronously pushes alarm information to team maintenance personnel, safety management personnel, duty plant managers, the superior dispatch center, and the local emergency management platform, along with captured images, reservoir water level monitoring data, and BIM evacuation drawings.

[0083] The hash evidence storage unit generates a unique hash value for all data in the process and stores it in a distributed manner. The work order closed-loop unit completes event archiving. The strategy self-optimization unit automatically optimizes the warehouse intrusion identification alarm parameters based on the valid alarm samples.

[0084] Through the synergistic effect of the above-mentioned technical features, the present invention achieves the following technical effects: the multi-source composite alarm verification mechanism uses video AI feature recognition and dedicated sensor timing joint verification for dual verification. It determines whether multi-source signals are triggered synchronously within a preset time window, effectively filtering single-source environmental interference and overcoming the technical problem of high false alarm rate of single sensor in complex hydropower plant environments.

[0085] The combination of six types of risk scenario-specific verification rules and a three-level differentiated linkage strategy library enables full-scenario coverage and precise graded response to multiple types of risks across the entire hydropower plant, overcoming the technical problems of incomplete scenario coverage and rigid linkage logic in the existing fragmented security system.

[0086] The BIM dynamic evacuation submodule integrates 3D BIM drawings and UWB real-time personnel positioning data, avoids risk areas in real time, and generates evacuation routes based on corridor topology, overcoming the technical problem of not being able to adjust the route in real time when relying on static escape diagrams.

[0087] The onCall hierarchical push submodule accurately matches the push level according to the alarm level and distributes alarm information synchronously through multiple channels, overcoming the technical problems of information overload and high-risk alarm omission. The hash evidence storage unit generates a unique hash value for the entire data chain and stores it in a distributed manner to achieve tamper-proof evidence storage and meet the compliance requirements for data integrity in the power industry's security audit.

[0088] The strategy self-optimization unit uses a Bayesian model to iteratively update the verification parameters, enabling the system to autonomously iteratively optimize itself and overcoming the technical problem that adjusting the false alarm threshold relies entirely on manual on-site debugging. The above modules work together to achieve a complete closed loop of sensing and data acquisition, composite verification, hierarchical linkage, dynamic evacuation, reliable evidence storage, and self-optimization.

[0089] It is understood that the specific implementation methods of the aforementioned multi-source composite alarm verification, hierarchical linkage strategy, BIM dynamic evacuation, and Bayesian self-optimization are not limited to those described above. The multi-source composite alarm verification can also be expanded or adjusted in terms of the number of verification sources for each scenario based on the sensor configuration of a specific hydropower plant. The confidence score range division of the three-level linkage strategy library can be dynamically adjusted according to the risk level standards of a specific hydropower plant. The BIM dynamic evacuation submodule can also use other indoor positioning technologies to replace UWB positioning. The update trigger cycle of the Bayesian model can be dynamically adjusted according to the actual closed-loop work order accumulation speed and operation and maintenance needs. The above parameters and implementation methods can be flexibly configured according to specific application scenarios and deployment conditions. All the above variations are within the scope of protection of this application.

Claims

1. A multi-source composite verification system for hydropower plants, characterized in that, include: The multi-source heterogeneous sensing and acquisition module connects to the heterogeneous sensing devices in the hydropower plant through an industrial protocol gateway, collects multiple types of sensing data, and connects to the business data system to output the data after unified parsing of the multi-source data. The multi-source composite alarm verification module performs time-series joint verification of video AI feature recognition and corresponding category dedicated sensors for multiple preset risk scenarios. When multiple source signals are triggered synchronously within a preset time window, it outputs an alarm signal containing the risk scenario, risk coordinates and confidence score. When a single source is triggered, it is determined to be environmental interference and filtered out. The hydropower scenario-based hierarchical linkage strategy scheduling module has a pre-stored hierarchical linkage strategy library corresponding to the various risk scenarios. Based on the received alarm signals, it matches the corresponding scenario and level, generates the corresponding linkage control command, and sends it to the on-site execution equipment. The BIM dynamic evacuation and onCall hierarchical push module dynamically avoids risk areas and generates evacuation routes when an alarm is triggered. At the same time, it matches the push level according to the alarm level and distributes multimodal alarm information synchronously through multiple channels. The alarm closed-loop trusted evidence storage self-optimization module generates a unique hash value for the entire link data and stores it in a distributed manner; at the same time, it automatically generates an electronic processing work order to complete the closed-loop archiving in response to alarms, and uses a machine learning model to continuously iterate and optimize the verification time window parameters and confidence level boundary values ​​for each scenario based on historical closed-loop samples.

2. The hydropower plant scenario-based and hierarchical security linkage system with full-domain multi-source composite verification as described in claim 1, characterized in that: The preset time window is 3 to 10 seconds, and the time window is dynamically adjusted according to the Bayesian model update results of the strategy self-optimization unit.

3. The hydropower plant scenario-based and hierarchical security linkage system with full-domain multi-source composite verification as described in claim 1, characterized in that: The time-series joint verification rules for each risk scenario in the multi-source composite alarm verification module are as follows: In a fire scenario, an effective alarm is output when at least two of the signals from the industrial TV's flame and smoke AI recognition system are triggered simultaneously with the signals from the smoke detector, heat sensor, and CO concentration sensor. In an SF6 leak scenario, an effective alarm is output when the thermal imaging gas cloud identification signal and the SF6 infrared concentration sensor signal are triggered synchronously. In the scenario of a flooded factory, a valid alarm is output when the video water accumulation recognition signal is triggered synchronously with the signals from at least two liquid level sensors. In the case of high voltage over-limit scenario at the booster station, an effective alarm will be output when the human body AI recognition signal, millimeter-wave radar signal and thermal imaging temperature measurement signal are triggered simultaneously. In the scenario of intrusion into the reservoir area, an effective alarm is output when the AI ​​recognition signal of a vessel or person is triggered synchronously with the shoreline radar signal. When construction workers cross the boundary, a valid alarm is output when the facial recognition signal, RFID positioning signal, and maintenance work order authorization signal are triggered simultaneously.

4. The hydropower plant scenario-based and hierarchical security linkage system with full-domain multi-source composite verification as described in claim 1, characterized in that, The hierarchical linkage strategy library is divided into the following categories according to confidence scores: A confidence score in the first preset interval corresponds to Level 1 risk, a confidence score in the second preset interval corresponds to Level 2 risk, and a confidence score in the third preset interval corresponds to Level 3 risk. In response to fire scenarios, Level 1 involves activating local ventilation and issuing a pop-up warning on the central control platform; Level 2 involves adding zoned smoke extraction, unlocking all access control doors in the corresponding fire compartment, forcing elevators to land, and activating emergency broadcasts; Level 3 involves adding automatic power cut-off to the high-voltage power supply to the fire area, activating automatic fire extinguishing devices, setting all underground access control doors and warehouse gates to always open, dynamically switching to backup evacuation routes, and pushing alarms and factory structure data to the higher-level dispatch platform and local emergency platform.

5. The hydropower plant scenario-based and hierarchical security linkage system with full-domain multi-source composite verification as described in claim 1, characterized in that: The BIM dynamic evacuation and onCall hierarchical push module includes a BIM dynamic evacuation submodule and an onCall hierarchical push submodule. The BIM dynamic evacuation submodule connects to the factory's BIM drawings and real-time personnel location data. When a preset level alarm is triggered, it avoids risk areas with risk coordinates as the center and generates evacuation routes based on corridor topology. When the risk range expands, it switches to a backup channel. For scenarios involving narrow underground corridors, the BIM dynamic evacuation submodule unlocks access control doors sequentially along the corridor topology path and simultaneously generates evacuation routes in both upstream and downstream directions; for scenarios involving multi-story factory buildings, it generates vertical evacuation routes excluding the floors where the fire originated.

6. The hydropower plant scenario-based and hierarchical security linkage system with full-domain multi-source composite verification as described in claim 1, characterized in that: The onCall hierarchical push submodule matches the corresponding push level according to the alarm level and distributes alarm information synchronously through multiple channels. The onCall hierarchical push submodule has the following push levels: Level 1 alarms are pushed to the team maintenance personnel; Level 2 alarms are pushed to the team maintenance personnel and safety supervision personnel; Level 3 alarms are pushed to the team maintenance personnel, safety supervision personnel, duty plant manager, superior dispatch center and local emergency management platform simultaneously; the push attachments include alarm coordinates, real-time video snapshots, BIM evacuation drawings and monitoring data curves.

7. The hydropower plant scenario-based and hierarchical security linkage system with full-domain multi-source composite verification as described in claim 1, characterized in that: The alarm closed-loop trusted evidence storage self-optimization module includes a hash evidence storage unit, a work order closed-loop unit, and a policy self-optimization unit; the hash evidence storage unit generates unique hash values ​​for each of the entire link data and stores them in a distributed manner; the work order closed-loop unit automatically generates electronic processing work orders in response to alarm signals and receives processing data uploaded by maintenance personnel to complete closed-loop archiving. The strategy self-optimization unit reads historical samples of closed-loop work orders and uses a machine learning model to iteratively update the composite verification time window parameters and confidence level boundary values ​​for each risk scenario. In the strategy self-optimization unit, the input features of the Bayesian model include the triggering time difference of each sensor, ambient temperature and humidity, and light intensity. The prior distribution uses historical false alarm rate statistics. The model is updated once for every preset number of closed-loop work orders.

8. The hydropower plant scenario-based and hierarchical security linkage system with full-domain multi-source composite verification according to claim 1, characterized in that, The system is also equipped with data interaction interfaces with the superior dispatch platform and the local emergency platform, which are used to synchronously upload alarm event information, plant BIM structure data and real-time on-site monitoring data to the superior dispatch platform and the local emergency platform.

9. The hydropower plant scenario-based and hierarchical security linkage system with full-domain multi-source composite verification as described in claim 1, characterized in that: The full-domain multi-source heterogeneous sensing and acquisition module connects to the heterogeneous sensing devices through a multi-industry protocol gateway compatible with Modbus, ONVIF, MQTT, and Power 104 protocols. The acquisition range includes smoke detectors, temperature detectors, CO concentration sensors, SF6 gas sensors, liquid level sensors, industrial video and UWB personnel positioning data in the underground plant area; millimeter-wave perimeter radar, thermal imaging temperature measurement and high-voltage switch status data in the booster station area; and shoreline photoelectric cameras, ship monitoring radar and water level data in the reservoir area.

10. A method for multi-source composite verification of security linkage in hydropower plants across different scenarios and levels, characterized in that: Includes the following steps: S1. Access the heterogeneous sensing devices across the entire hydropower plant through the industrial protocol gateway, collect multiple types of sensing data, and connect to the business data system to complete the unified parsing of multi-source data; S2. For the preset multiple risk scenarios, perform video AI feature recognition and corresponding dedicated sensor time-series joint verification within the preset time window. Single-source triggers are judged as environmental interference and filtered out, while multi-source synchronous triggers are judged as valid alarms. An alarm signal containing risk scenario, risk coordinates and confidence score is output. S3. Match the pre-stored hierarchical linkage strategy library according to the risk scenario and confidence score of the alarm signal, generate the corresponding level of linkage control command and send it to the on-site execution device. S4. When a preset alarm level is triggered, based on the factory BIM drawings and real-time personnel location data, an evacuation route is generated to avoid the risk area with the risk coordinates as the center. When the risk range expands, the backup channel is switched. At the same time, the corresponding push level is matched according to the alarm level, and the alarm information is distributed synchronously through multiple channels. S5. Generate unique hash values ​​for each piece of data across the entire chain and store them in a distributed manner; automatically generate electronic processing work orders and complete closed-loop archiving after receiving processing data uploaded by maintenance personnel; S6. Read historical samples of closed-loop work orders and use machine learning models to iteratively update the composite verification time window parameters and confidence level boundary values ​​for each risk scenario.

Images