Thermal power plant 5G private network and intelligent perception fused safety inspection system

Abstract

The invention relates to the technical field of intelligent inspection, in particular to a thermal power plant 5G private network and intelligent perception fused safety inspection system, which comprises a 5G private network communication module used for constructing a low-delay and high-reliability localized network in a thermal power plant; the multi-source intelligent sensing terminal is used for acquiring equipment state and environment data in real time; the edge computing node is used for data preprocessing and real-time analysis; the cloud intelligent analysis platform is used for realizing equipment fault prediction and safety early warning; through cooperative work of the multi-source intelligent sensing terminals and in combination with local preprocessing capability of the edge computing nodes, real-time acquisition and preliminary analysis of equipment state and environment data are realized, redundant data transmission is reduced, fault detection efficiency is improved, meanwhile, a 5G private network provides stable localized communication guarantee for a thermal power plant inspection system, and the power plant inspection system has a wide application prospect. The problem of signal delay or interruption caused by interference of a traditional wireless network is avoided, and the response speed and reliability of the system are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent inspection, and particularly relates to a safety inspection system combining 5G private network and intelligent sensing in a thermal power plant. BACKGROUND

[0002] As an important energy infrastructure, the equipment operation state of a thermal power plant is directly related to the safety and stability of power supply. The existing thermal power plants generally adopt a combination of staff inspection and fixed sensor monitoring to periodically check key equipment such as boilers, steam turbines and coal conveying systems. Some advanced power plants have tried to introduce robot inspection and wireless sensor network technology to improve inspection efficiency and coverage.

[0003] The existing inspection system has poor coordination capability of multi-source sensing data and communication network. The traditional wireless network is easily affected by electromagnetic interference in the complex environment of a thermal power plant, resulting in high delay of inspection robot control instructions and unstable sensor data return, which makes it difficult to meet the real-time requirements. At the same time, the data collected by various intelligent sensing terminals lack an efficient fusion mechanism, and the edge computing capability is insufficient, resulting in low fault warning accuracy, frequent false positives and false negatives, and seriously affecting the operation and maintenance efficiency.

[0004] Therefore, in view of the above problems, the present application provides a safety inspection system combining 5G private network and intelligent sensing in a thermal power plant. By constructing a localized 5G network with high reliability and low latency, combining edge computing and cloud intelligent analysis, the precise sensing, real-time transmission and intelligent diagnosis of equipment state are realized, thereby improving the automation level and fault warning capability of safety inspection in a thermal power plant. SUMMARY

[0005] In order to overcome the problem of poor real-time performance of the existing system, the present application provides a safety inspection system combining 5G private network and intelligent sensing in a thermal power plant.

[0006] The technical solution of the present application is as follows: a safety inspection system combining 5G private network and intelligent sensing in a thermal power plant, comprising: a 5G private network communication module for constructing a localized network with low latency and high reliability in a thermal power plant; a multi-source intelligent sensing terminal including an infrared thermal imager, an acoustic sensor array, a vibration sensor, a gas sensor and a high-definition camera for real-time collection of equipment state and environmental data; an edge computing node deployed in key areas of the plant for data preprocessing and real-time analysis; a cloud intelligent analysis platform receiving data through the 5G private network to realize equipment fault prediction and safety warning.

[0007] As preferred, the 5G private network uses network slicing technology and is divided into inspection control slice, data backhaul slice and emergency communication slice. The inspection control slice is dedicated to ensuring the millisecond-level low-latency transmission of inspection robot and unmanned aerial vehicle control instructions. The data backhaul slice allocates differentiated bandwidth according to the priority of sensor data. The emergency communication slice is activated in the event of power interruption or fire and maintains key communication through pre-configured redundant channels. Virtual isolation technology is used between slices.

[0008] As preferred, the intelligent sensing terminal is built-in with an AI chip, which identifies abnormal overheating areas of equipment by comparing infrared thermal imaging data with historical temperature curves, judges the micro leakage points of boiler pipelines using voiceprint signals collected by an acoustic sensor array combined with a convolutional neural network, and integrates a gas sensor and a machine learning model to trigger a combustion and explosion risk level evaluation algorithm when detecting that the concentration of dangerous gas exceeds the standard, and then transmits the data to the cloud after local data compression through the 5G private network.

[0009] As preferred, the system further includes an autonomous inspection robot equipped with a 5G and UWB fusion positioning module, which automatically switches to UWB base station positioning in a plant with shielded satellite signals. The end of the robot's mechanical arm is equipped with a tool head that can be quickly replaced, and the robot receives operation instruction packages from the cloud through the 5G network. The internal cooling system of the robot can work continuously for more than 2 hours in a 150℃ environment.

[0010] As preferred, the edge computing node deploys a lightweight fault diagnosis model, in which an LSTM model analyzes steam turbine bearing vibration time series data at intervals of 10ms to predict abnormal vibration trends 30 minutes in advance, and a YOLOv5 model performs real-time identification on pressure gauge images captured by a camera with an error of less than ±0.5% range. The edge node synchronizes model parameter updates with the cloud platform once every 6 hours through the 5G private network, and enables local cache mode to maintain basic analysis functions when the network is interrupted.

[0011] As preferred, the cloud intelligent analysis platform constructs a three-dimensional virtual mapping of the thermal power plant through a BIM model, drives twin body state updates in real time by accessing sensor data returned by 5G, simulates the diffusion path of high-temperature steam after the steam pipeline is broken using fluid dynamics simulation, and generates an optimal inspection route that avoids dangerous areas based on the Dijkstra algorithm. Inspection personnel can view route navigation arrows and equipment health scores superimposed in the real scene through the AR interface of a mobile terminal.

[0012] As preferred, the system is provided with a security protection mechanism, in which the 5G private network uses quantum key distribution technology to encrypt the control instructions and sensor data of the robot, the key update frequency reaches 1 time / minute, the intelligent sensing terminal shell is embedded with a pressure sensitive film, the Flash memory data is automatically erased when illegal disassembly is detected, and the cloud platform writes the inspection records and alarm event data into the Ethereum test chain in a time stamp manner through a block chain storage module.

[0013] As preferred, the scheduling method of the autonomous inspection robot dynamically adjusts the task queue based on the real-time risk assessment results of the equipment, automatically increases the inspection frequency of a certain area to once every 15 minutes when the temperature sensor of the area exceeds the threshold value for 3 times in a row, and sends an electronic access card request to the factory access control system through the 5G private network when a sudden alarm occurs, the central controller coordinates the task handover of multiple robots, and triggers the charging instruction when the remaining power is 15%, and starts the standby robot within 500 meters to take over the task.

[0014] As preferred, the system comprises an AR assisted inspection function, which superimposes the cloud device archives in real time to the field of view of the inspection personnel AR glasses through the 5G private network, supports remote labeling of abnormal equipment areas by experts and generates three-dimensional arrow directions, and the three-dimensional disassembly diagram of the equipment can be called up by gesture operation of the inspection personnel, the AR system synchronously records the voice remarks and key attention area screenshots in the inspection process, and is automatically associated to the work order system of the intelligent digital analysis platform.

[0015] As preferred, the system further comprises an adaptive learning module, which optimizes the multi-sensor data fusion weight coefficient by analyzing historical false alarm data, automatically triggers online incremental training of the fault prediction model when the feedback of the maintenance work order shows that the actual fault of a fan bearing does not match the prediction result, and adjusts the transmission power of the 5G micro base station in the factory based on the signal strength thermal map of the inspection robot.

[0016] The beneficial effects of the present application are: 1. Through the cooperative work of the multi-source intelligent sensing terminal such as the infrared thermal imager, acoustic sensor array, gas sensor, vibration sensor and high-definition camera, combined with the local preprocessing capability of the edge computing node, real-time collection and preliminary analysis of equipment state and environmental data are realized, redundant data transmission is reduced, fault detection efficiency is improved, at the same time, the 5G private network provides stable localized communication guarantee for the power plant inspection system, ensures the millisecond-level transmission of the robot control instructions and sensor data, avoids the signal delay or interruption problem caused by traditional wireless network interference, and significantly improves the system response speed and reliability.

[0017] 2.The cloud intelligent analysis platform is based on high-quality data of 5G private network backhaul, uses big data analysis and artificial intelligence algorithm to realize early identification of equipment abnormal state and fault trend prediction, effectively reduces false positive rate and false negative rate, thereby improving the safety and economy of the operation and maintenance of the thermal power plant, the system adopts modular design, can flexibly deploy intelligent sensing terminals and edge computing nodes according to the inspection needs of different areas of the thermal power plant, and adapt to diversified business scenarios through the network slicing technology of the 5G private network, and meet the customized needs in complex industrial environments. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 A system framework schematic diagram of the present application is shown. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical scheme and advantages of the embodiments of the present application more clear, the technical scheme in the embodiments of the present application will be described clearly and completely below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application.

[0020] Please refer to Figure 1 The present application provides an embodiment: a safety inspection system of thermal power plant 5G private network and intelligent sensing fusion, comprising: A 5G private network communication module is used to build a localized network with low latency and high reliability in the thermal power plant; A multi-source intelligent sensing terminal includes an infrared thermal imager, an acoustic sensor array, a vibration sensor, a gas sensor and a high-definition camera, which can collect device state and environmental data in real time; An edge computing node is deployed in the key area of the plant for data preprocessing and real-time analysis; A cloud intelligent analysis platform receives data through the 5G private network to realize device fault prediction and safety warning.

[0021] Further, the system first collects multi-dimensional data such as temperature, gas concentration, mechanical vibration and visual state of the equipment in real time through the multi-source intelligent sensing terminal deployed in the key area of the thermal power plant. These data are transmitted to the edge computing node deployed nearby through the low-latency and high-reliability localized network built by the 5G private network communication module for data preprocessing, including noise filtering, feature extraction and preliminary abnormality judgment. The processed key data are further uploaded to the cloud intelligent analysis platform through the 5G private network. The platform realizes deep analysis and prediction of equipment failure by combining historical data and machine learning model, and finally generates an inspection report or a warning instruction and feeds back to the terminal equipment or the operation and maintenance personnel. The application guarantees the real-time transmission reliability of massive sensing data through the 5G private network, effectively reduces the cloud processing load and shortens the response time through the edge computing node, realizes the stereoscopic monitoring of the equipment state through the cooperative work of the multi-source sensing terminal, and significantly improves the fault diagnosis accuracy through the intelligent analysis of the cloud platform. The whole system has made a breakthrough improvement in communication real-time, data fusion depth and operation efficiency compared with the traditional inspection method.

[0022] The 5G private network uses network slicing technology and is divided into an inspection control slice, a data backhaul slice and an emergency communication slice. The inspection control slice is used to guarantee the millisecond-level low-latency transmission of the control instructions of the inspection robot and the unmanned aerial vehicle. The data backhaul slice allocates differentiated bandwidth according to the priority of the sensor data. The emergency communication slice is automatically activated in the event of power interruption or fire and the like, maintains key communication through a pre-configured redundant channel, and uses virtual isolation technology between the slices.

[0023] Further, the inspection control slice is used to transmit the motion control instructions and real-time feedback data of the inspection robot and the unmanned aerial vehicle. The QoS guarantee mechanism with ultra-low latency is used to realize the instruction transmission delay <10ms, so as to ensure the accurate control and emergency braking of the equipment. The data backhaul slice preferentially transmits high-priority sensor data such as high-temperature alarm and gas leakage through a dynamic bandwidth allocation strategy, and simultaneously uses data compression and differential transmission technology to reduce the transmission bandwidth of the regular monitoring data by 40%, thereby optimizing the resource utilization of the network while guaranteeing the real-time of the key data. The emergency communication slice is automatically activated when an abnormal event such as power interruption or fire is detected, starts the pre-set backup communication link and anti-interference transmission protocol, and still can maintain the transmission of the video monitoring data and alarm signals in the core area under the condition that the backbone network is damaged, thereby ensuring the emergency response timeliness.

[0024] The three slices realize the sharing but logical isolation of the physical resources through the virtualization technology, meet the differentiated needs of different business scenarios, and improve the overall network resource utilization efficiency, so that the system can still maintain stable communication performance in a complex industrial environment.

[0025] The intelligent sensing terminal is internally provided with an AI chip, which identifies the overheating abnormal area of the equipment through the comparison of the infrared thermal imaging data and the historical temperature curve, judges the micro leakage point of the boiler pipeline by using the voiceprint signals collected by the acoustic sensor array and the convolutional neural network, simultaneously integrates the gas sensor and the machine learning model, automatically triggers the combustion and explosion risk level evaluation algorithm when the concentration of the dangerous gas exceeds the standard, and transmits the data to the cloud after local data compression through the 5G private network.

[0026] Further, after the infrared thermal imager collects the surface temperature distribution data of the equipment, the AI chip compares it with the normal operation temperature threshold library in real time, and when it detects that the local temperature exceeds the safety threshold of 50 DEG C, it immediately marks the overheating area and triggers an alarm, the acoustic sensor array continuously collects the equipment operation noise, analyzes the voiceprint features through the pre-trained convolutional neural network model, can identify the specific frequency sound wave generated by the 0.1mm level micro leakage of the boiler pipeline, the gas sensor monitors the concentration of CO, H2 and other dangerous gases, combined with the LSTM time series prediction model, when the gas concentration gradient change rate is abnormal, the deflagration risk assessment algorithm is automatically started, all sensing data completes feature extraction and compression processing locally in the terminal, only the key feature data is uploaded through the 5G private network, compared with the original data transmission volume, it is reduced by 70%.

[0027] The present application realizes real-time diagnosis of equipment state through edge intelligent processing, improves the minute-level response of traditional cloud analysis to millisecond level, and greatly reduces the network transmission load, and is particularly suitable for early fault accurate detection in high temperature and high noise environment of thermal power plant.

[0028] Further, through the 5G and UWB fusion positioning module, the plant is automatically switched to the UWB base station positioning mode in the poor satellite signal coverage, realizes the real-time position tracking with the accuracy of ±3cm, the multifunctional tool head carried by the mechanical arm can automatically replace the adapter according to the instruction issued by the cloud, and perform diversified tasks such as infrared temperature measurement and valve operation, the explosion-proof cooling system ensures that the robot continuously and stably operates for more than 2 hours in the 150 DEG C high temperature environment, the robot transmits the inspection data in real time through the 5G private network and receives the control instruction, and when detecting the equipment abnormity, automatically adjusts the inspection path to prioritize the troubleshooting point.

[0029] The edge computing node deploys a lightweight fault diagnosis model, wherein the LSTM model analyzes the steam turbine bearing vibration time series data at an interval of 10ms, predicts the abnormal vibration trend 30 minutes in advance, the YOLOv5 model performs real-time identification on the pressure gauge image shot by the camera, the error is less than ±0.5% range, and the edge node synchronizes the model parameter update with the cloud platform through the 5G private network every 6 hours, and enables the local cache mode to maintain the basic analysis function when the network is interrupted.

[0030] Further, the edge computing nodes deployed in the key areas of the thermal power plant first receive raw data streams from various intelligent sensing terminals, including turbine bearing vibration waveforms sampled at 10ms intervals, pressure gauge images captured by high-definition cameras, and temperature distribution data collected by infrared thermal imagers. For vibration data, the built-in LSTM time series analysis model extracts real-time time-frequency domain features of the vibration signal. By comparing with the feature library under the historical normal operating state, an early warning can be given 30 minutes in advance when early signs of bearing wear appear. The detection sensitivity reaches the level of 0.01mm of micro-displacement. For pressure gauge images, the optimized YOLOv5 model completes real-time recognition on the edge, converting the positions of the gauge pointers to accurate pressure readings with an identification error controlled within ±0.5% of the range. Meanwhile, the edge computing nodes run lightweight data fusion algorithms to perform feature-level fusion on the correlation data from multiple sensors, such as combining vibration anomalies with temperature rise data to comprehensively judge the equipment health status, effectively reducing the false alarm rate of a single sensor. All processing results are uploaded in real time to the cloud intelligent analysis platform through the 5G private network. The edge nodes automatically synchronize the latest model parameters with the cloud every 6 hours. In the event of network interruption, the latest locally cached models are automatically enabled to continue providing services. The nodes use a hierarchical storage strategy, with raw data retained for 24 hours and feature data retained for 7 days, and support remote triggering of reanalysis of specific time period data.

[0031] The cloud intelligent analysis platform constructs a three-dimensional virtual mapping of the thermal power plant through a BIM model, drives the twin state update using real-time sensor data returned through 5G, simulates the high-temperature steam diffusion path after a steam pipe rupture using fluid dynamics simulation, and generates an optimal inspection route that avoids dangerous areas based on the Dijkstra algorithm. Inspection personnel can view the route navigation arrows and equipment health scores superimposed in the real scene through the mobile terminal AR interface.

[0032] Further, a three-dimensional virtual model of the entire thermal power plant is first constructed based on a BIM model. Real-time access is made to equipment temperature, pressure, and vibration parameters uploaded by edge computing nodes through a 5G private network, driving the digital twin and physical equipment to keep synchronized updates. The platform uses a fluid dynamics simulation engine to automatically simulate the high-temperature steam leakage diffusion path, predict the affected area range, and generate a device health score and visual display when an abnormal pressure in a steam pipe is detected. When an alarm event is triggered, the platform dynamically generates an optimal inspection path based on a comprehensive analysis of equipment location, dangerous area distribution, and real-time location of inspection robots. This path information is pushed to inspection personnel AR terminals through a 5G network and displayed in the actual plant environment in the form of three-dimensional arrows. Meanwhile, the platform automatically associates maintenance records and spare parts inventory information for faulty equipment to generate an electronic work order containing disposal suggestions.

[0033] The system is provided with a security protection mechanism, which uses quantum key distribution technology to encrypt the control instructions and sensor data of the robot, with a key update frequency of 1 time / minute. The intelligent sensing terminal shell is embedded with a pressure-sensitive film, which automatically erases the Flash memory data when illegal disassembly is detected. The cloud platform writes the inspection records and alarm event data into the Ethereum test chain in a time-stamped manner through the block chain storage module.

[0034] The scheduling method of the autonomous inspection robot dynamically adjusts the task queue based on the real-time risk assessment results of the equipment. When the temperature sensor in a certain area exceeds the threshold for 3 consecutive times, the inspection frequency of that area is automatically increased to 1 time every 15 minutes. In the event of a sudden alarm, the robot sends an electronic access card request to the factory access control system through the 5G private network. The central controller coordinates the task handover of multiple robots and triggers the charging instruction when the remaining power is 15%. At the same time, it starts the standby robot within 500 meters to take over the task and ensures that the inspection coverage rate does not decrease.

[0035] Further, the system receives real-time equipment state data from multiple source intelligent sensing terminals and edge computing nodes, dynamically calculates the risk level of each area based on the preset risk assessment model (considering parameters such as temperature exceeding frequency, vibration amplitude, gas concentration, etc.), and automatically increases the inspection frequency of the area from the regular 2 hours / time to 15 minutes / time when the temperature sensor in the area detects the threshold for 3 consecutive times. When a sudden high temperature or gas leakage alarm is triggered, the adjacent robot immediately sends an electronic access card request containing a digital signature to the factory access control system through the 5G private network. After obtaining access rights within 200ms, it autonomously plans the shortest path to avoid obstacles based on fused positioning data. During the journey, it continuously transmits real-time video and sensor data back through the 5G network. The intelligent power management system built into the robot monitors the power state in real time and automatically sends a charging request to the control center when the remaining power is less than 15%. The system immediately dispatches a standby robot within a 500-meter range to take over, and ensures that the inspection coverage rate remains at 100% during the task handover period through the cloud-based intelligent analysis platform.

[0036] The system includes an AR-assisted inspection function that superimposes real-time cloud device archives onto the inspection personnel's AR glasses field of view through the 5G private network, supports remote labeling of abnormal device areas by experts and generates three-dimensional arrow pointers. Inspection personnel can retrieve device three-dimensional disassembly diagrams through hand gestures. The AR system synchronously records voice notes and key attention area screenshots during the inspection process and automatically associates them to the intelligent analysis platform's work order system.

[0037] The system also comprises an adaptive learning module that optimizes multi-sensor data fusion weight coefficients by analyzing historical false alarm data, automatically triggers online incremental training of the fault prediction model when the maintenance work order feedback that the actual fault of a fan bearing does not match the prediction result, and adjusts the transmission power of the 5G micro base station in the plant based on the signal strength heat map of the inspection robot.

[0038] Further, the workflow of the present application is described in detail as follows: The present application first collects equipment state data in real time through multi-source intelligent sensing terminals deployed in key areas of the thermal power plant, including continuous monitoring of equipment surface temperature by temperature sensors, acoustic fingerprint signals collected by acoustic sensor arrays, detection of hazardous gas concentration by gas sensors, recording of mechanical operation parameters by vibration sensors, and transmission of these raw data to edge computing nodes for preliminary processing and analysis.

[0039] After receiving the sensing data, the edge computing node immediately runs the built-in LSTM time series analysis model to process the vibration data and identify early fault features, and analyzes the pressure gauge image through the YOLOv5 model, all analysis results are completed within 200ms and uploaded to the cloud intelligent analysis platform, significantly reducing the delay of traditional cloud processing.

[0040] The cloud intelligent analysis platform updates the three-dimensional virtual model based on the real-time uploaded data, automatically starts fluid dynamics simulation when detecting abnormal parameters, predicts the fault impact range, and evaluates the equipment health status combined with historical data, generates visual reports and disposal suggestions.

[0041] The system dynamically assesses the risk level of each area based on the analysis results of the cloud intelligent analysis platform, automatically adjusts the inspection frequency of the area when the temperature of the area continuously exceeds the standard, and sends new task instructions to the inspection robot through the 5G private network.

[0042] The inspection robot receives the instructions and immediately plans the optimal path, continuously transmits real-time monitoring data through the 5G network during the journey, and automatically requests a backup robot to take over when the power is insufficient, thereby ensuring the continuity and full coverage of the inspection work.

[0043] The entire system realizes high-speed data transmission and collaborative control between components through the 5G private network, improves the fault identification efficiency, and improves the speed of emergency response, while supporting simultaneous scheduling of up to 20 robots, fully meeting the intelligent inspection needs of large thermal power plants.

[0044] Further, the present application provides an embodiment: deploy the system in the boiler room of a certain 1000 MW coal-fired power plant, install 12 intelligent sensing terminals (including infrared, vibration, gas sensors) and 3 edge computing nodes, realize data interaction through 5G private network, when the boiler water-cooled wall pipe appears 2mm crack, the vibration sensor detects abnormal spectrum characteristics within 30 seconds, the edge node LSTM model predicts the pipe wall rupture risk 42 minutes in advance, the system automatically schedules the nearest inspection robot to confirm, and shows the crack position three-dimensional label to the operation and maintenance personnel through the AR terminal, so that the repair response time is shortened from the conventional 4 hours to 1.5 hours, avoiding the non-planned shutdown loss of about 800,000 yuan.

[0045] Further, the present application provides an embodiment: applied to a coastal power plant steam turbine workshop, the sensor protection level is particularly optimized for high-salt mist environment, when the typhoon weather causes the workshop to be locally flooded, the emergency communication slice is automatically activated, the 2 robots in the flooded area are switched to the waterproof mode to continue working, the water level data and equipment insulation state are real-time returned through the 5G network, the command center remotely closes the affected unit and starts the drainage plan according to the water immersion influence range simulated by the cloud intelligent analysis platform, compared with the traditional manual inspection method, the emergency treatment is completed 3 hours in advance, and the core components of the steam turbine worth 20 million yuan are protected from corrosion damage.

Claims

1. A thermal power plant 5G private network and intelligent sensing fusion safety inspection system, characterized in that, The system comprises: 5G private network communication module, used for building a localized network with low latency and high reliability in the power plant; Multi-source intelligent sensing terminal, including infrared thermal imager, acoustic sensor array, vibration sensor, gas sensor and high-definition camera, for real-time collection of equipment state and environmental data; Edge computing node, deployed in key areas of the plant, for data preprocessing and real-time analysis; Cloud intelligent analysis platform, receiving data through 5G private network, for equipment fault prediction and safety warning.

2. The power plant 5G private network and intelligent sensing integrated security inspection system according to claim 1, characterized in that: The 5G private network uses network slicing technology, which is divided into inspection control slice, data backhaul slice and emergency communication slice. The inspection control slice is used to ensure the transmission of millisecond-level low latency control instructions of inspection robots and drones. The data backhaul slice allocates differentiated bandwidth according to the priority of sensor data. The emergency communication slice is activated in the event of power interruption or fire, etc. to maintain key communication through pre-configured redundant channels, and virtual isolation technology is used between slices.

3. The power plant 5G private network and intelligent sensing integrated security inspection system according to claim 1, characterized in that: The multi-source intelligent sensing terminal is built-in AI chip, which can identify abnormal overheating areas of equipment by comparing infrared thermal imaging data with historical temperature curve, and can use acoustic sensor array to collect voiceprint signals and judge the micro leakage point of boiler pipeline through convolutional neural network. When the concentration of dangerous gas exceeds the standard, the gas sensor and machine learning model are integrated to trigger the explosion risk level evaluation algorithm, and the data is compressed locally and then transmitted to the cloud through the 5G private network.

4. The power plant 5G private network and intelligent sensing integrated security inspection system according to claim 1, characterized in that: The system also includes an autonomous inspection robot equipped with a 5G and UWB fusion positioning module. In the plant where satellite signals are shielded, it automatically switches to UWB base station positioning. The mechanical arm end is equipped with a tool head that can be quickly replaced, and it receives operation instruction packages from the cloud through the 5G network. The internal cooling system of the robot can work continuously for more than 2 hours in a 150℃ environment.

5. The power plant 5G private network and intelligent sensing integrated security inspection system according to claim 1, characterized in that: The edge computing node deploys a lightweight fault diagnosis model. The LSTM model analyzes turbine bearing vibration time series data at 10ms intervals to predict abnormal vibration trends 30 minutes in advance. The YOLOv5 model performs real-time identification on the pressure gauge image captured by the camera, with an error of less than ±0.5% range. The edge node synchronizes model parameter updates with the cloud platform once every 6 hours through the 5G private network, and enables local cache mode to maintain basic analysis functions when the network is interrupted.

6. The power plant 5G private network and intelligent sensing integrated security inspection system according to claim 1, characterized in that: The cloud intelligent analysis platform builds a three-dimensional virtual mapping of the power plant through BIM model, drives twin body state update by real-time access to 5G backhaul sensor data, simulates the high-temperature steam diffusion path after the steam pipe breaks using fluid dynamics simulation, and generates an optimal inspection route that avoids dangerous areas based on Dijkstra algorithm. Inspection personnel can view route navigation arrows and equipment health scores superimposed on the real scene through the AR interface of the mobile terminal.

7. The power plant 5G private network and intelligent sensing integrated security inspection system according to claim 1, characterized in that: The system is provided with a security protection mechanism, in which the 5G private network uses quantum key distribution technology to encrypt the control instructions and sensor data of the robot, the key update frequency reaches 1 time / minute, the intelligent sensing terminal shell is embedded with a pressure sensitive film, which automatically erases the Flash memory data when illegal disassembly is detected, and the cloud platform writes the inspection records and alarm event data into the Ethereum test chain in the form of time stamp through the block chain storage module.

8. The power plant 5G private network and intelligent sensing integrated security inspection system according to claim 4, characterized in that: The scheduling method of the autonomous inspection robot dynamically adjusts the task queue based on the real-time risk assessment results of the equipment, automatically increases the inspection frequency of a certain area to once every 15 minutes when the temperature sensor of the area exceeds the threshold for 3 times in a row, and when a sudden alarm occurs, the robot sends an electronic access card request to the factory gate system through the 5G private network, the central controller coordinates the task handover of multiple robots, and triggers the charging instruction when the remaining power is 15%, and starts the standby robot within 500 meters to take over the task. 9.The power plant 5G private network and intelligent sensing fusion security inspection system according to claim 1, characterized in that: The system includes an AR assisted inspection function, which superimposes the real-time device file in the cloud to the field of view of the inspection personnel's AR glasses through the 5G private network, supports remote labeling of abnormal equipment areas by experts and generates three-dimensional arrow directions, and the inspection personnel can call up the three-dimensional disassembly diagram of the equipment through gesture operation, the AR system synchronously records the voice remarks and key attention area screenshots in the inspection process, and automatically associates them to the work order system of the intelligent digital analysis platform.

10. The power plant 5G private network and intelligent sensing integrated security inspection system according to claim 1, characterized in that: The system also includes an adaptive learning module, which optimizes the multi-sensor data fusion weight coefficient by analyzing historical false alarm data, automatically triggers online incremental training of the fault prediction model when the maintenance work order feedback that the actual fault of a fan bearing does not match the prediction result, and adjusts the transmission power of the 5G micro base station in the factory based on the signal strength thermal map of the inspection robot.

Images