Intelligent safety prevention and control system and method for hydropower station based on unattended operation

Abstract

The invention discloses an unattended hydropower station intelligent safety prevention and control system and method, and the system comprises a sensing layer, a network layer, a platform layer and an application layer, the sensing layer is provided with multiple types of sensors, and collects equipment operation parameters, a dam structure state, a reservoir area hydrological environment and perimeter image data in real time; the underwater sonar, the unmanned aerial vehicle and the camera are included to realize omnibearing monitoring; the network layer adopts 5G + LoRa dynamic fusion networking, and a deep Q network routing algorithm is integrated through an intelligent gateway. According to the intelligent safety prevention and control system and method for the hydropower station based on unattended operation, the structure is reasonable, a three-dimensional monitoring network is constructed through vibration / temperature / pressure sensors, underwater sonar, an unmanned aerial vehicle, an AI camera and the like, and equipment operation parameters, the dam structure state, the reservoir area hydrological environment and perimeter image data are collected in real time; all-around dead-corner-free monitoring of equipment, structure and environment of the hydropower station is achieved, and the system is worthy of popularization.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent safety control of hydropower stations, and in particular relates to an unmanned intelligent safety control system and method for hydropower stations. Background Art

[0002] As a core component of the global clean energy system, hydropower stations generated 4.5 trillion kilowatt-hours of electricity worldwide in 2023, accounting for 54% of all renewable energy generation. They play an irreplaceable role in optimizing the energy mix and addressing climate change. However, traditional hydropower stations rely heavily on manual oversight. Statistics show that a medium-sized hydropower station requires over 2,000 inspections annually, with a single inspection taking 8-12 hours. Furthermore, manual inspections have a 30% chance of missing hidden faults, such as localized overheating and minor cracks. Furthermore, the safety risks posed by complex environments cannot be ignored. Between 2022 and 2024, an average of 15 hydropower station equipment failures in China due to manual inspection oversight occurred annually, resulting in direct economic losses exceeding 200 million yuan.

[0003] In terms of monitoring technology, traditional hydropower stations mainly rely on single-point sensors and manual inspections, which makes it difficult to achieve full-scene coverage of equipment operating status, dam structural safety, reservoir hydrological environment and perimeter safety. For example, equipment fault monitoring often results in incomplete feature extraction due to the single sensor type, and manual inspections have problems such as long cycles and high missed detection rates. As a result, early abnormal signals such as bearing inner and outer ring failures are easily overlooked, and the risk of unplanned downtime is significantly increased. In addition, traditional video surveillance can only achieve real-time feedback of images and lacks the ability to intelligently identify perimeter intrusions and abnormal motion trajectories. Safety warnings rely on manual interpretation, and the problem of response lag is prominent.

[0004] Therefore, it is necessary to design corresponding technical solutions. Summary of the Invention

[0005] The present invention provides an unmanned intelligent safety control system and method for a hydropower station, which solves the problem.

[0006] To achieve the above objectives, the present invention provides the following technical solutions: an unmanned intelligent safety control system and method for hydropower stations, comprising a perception layer, a network layer, a platform layer, and an application layer. The perception layer deploys multiple types of sensors to collect real-time data on equipment operating parameters, dam structural status, reservoir hydrological environment, and perimeter images, and includes underwater sonar, drones, and cameras to achieve all-round monitoring;

[0007] The network layer adopts 5G+LoRa dynamic fusion networking, integrates the Deep Q network routing algorithm through the intelligent gateway, dynamically switches between 5G high-speed links and LoRa wide-area links based on link quality, and configures a satellite communication backup link;

[0008] The platform layer builds a digital twin-driven big data analysis platform, which uses AI algorithms such as LSTM and YOLOv5 to predict equipment failures and identify abnormal events, and achieves real-time bidirectional synchronization between virtual models and physical power plants.

[0009] The application layer supports hierarchical early warning, automatic triggering of emergency strategies, and three-dimensional visual remote operation and maintenance, and has multi-terminal access capabilities.

[0010] Preferably, the deep Q network routing algorithm of the network layer monitors signal strength, latency and packet loss rate in real time, giving priority to ensuring low-latency transmission of real-time data, while reducing power consumption of low-speed data transmission, and realizing intelligent path optimization of multi-network integration.

[0011] Preferably, the digital twin platform uses CFD fluid mechanics and structural mechanics models to simulate extreme scenarios such as flood impact and equipment fatigue, rehearse the impact of failures and optimize emergency strategies, forming a closed-loop control of physical equipment and virtual models.

[0012] Preferably, the security protection system includes a network layer firewall, data layer AES-256 encryption and application layer instruction secondary verification mechanism, combined with an AI threat detection model to block abnormal attacks in real time and switch to backup links.

[0013] Preferably, the method comprises the following steps: S1, a sensor collects device / environment data in real time, and generates a standardized data frame through noise reduction and feature extraction;

[0014] S2, selects the optimal path through 5G+LoRa fusion link and deep Q network algorithm, and switches to satellite communication to ensure reliable data upload in case of abnormality;

[0015] S3: Input data drives the synchronous update of the virtual model, and the AI ​​algorithm integrates analysis to generate equipment failure probability and environmental risk assessment results;

[0016] S4. Trigger audible and visual alarms, remote notifications, and automatic equipment control based on risk levels. Instructions are verified twice by edge nodes to prevent attacks.

[0017] Preferably, in the hierarchical linkage response, the red alert triggers a satellite SMS notification for downstream evacuation, and at the same time records the emergency process and feeds it back to the digital twin platform to achieve self-learning optimization of the emergency plan.

[0018] The beneficial effects of the present invention are as follows:

[0019] 1. This unmanned intelligent safety control system and method for hydropower stations uses vibration / temperature / pressure sensors, underwater sonar, drones, and AI cameras to build a three-dimensional monitoring network. This system collects real-time data on equipment operating parameters, dam structural status, reservoir hydrological environment, and perimeter imagery, enabling comprehensive, all-encompassing monitoring of hydropower station equipment, structures, and environments. It also uses FFT spectrum analysis to extract characteristic frequencies of equipment faults and, combined with the 3σ rule, filters out noise in real time, achieving a compression ratio of 10:1. This reduces data transmission pressure while accurately capturing fault characteristics of bearing inner and outer races, providing a highly reliable data foundation for equipment health assessments.

[0020] 2. This unmanned intelligent safety control system and method for hydropower stations is based on 5G+LoRa integrated networking and Deep Q network routing algorithm. It monitors signal strength, latency, packet loss rate and sensor power in real time, and prioritizes the transmission of real-time data such as video key frames and control instructions through 5G high-speed links. It transmits low-speed environmental data through LoRa wide-area links, realizing intelligent path optimization of "high-speed data with low latency and low-speed data with low power consumption". Satellite communication is configured as a backup link, and it automatically switches when the dual-network signal is below the threshold. Encoded key data is sent every 1 minute to ensure uninterrupted data transmission in extreme scenarios, solving the problem that traditional single-network transmission is easily restricted by the geographical environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a schematic diagram of the system architecture layer of the present invention;

[0022] Figure 2 This is the 5G+LoRa dynamic networking data transmission flow chart of the present invention;

[0023] Figure 3 This is a flow chart of the intelligent early warning and emergency response of the present invention. DETAILED DESCRIPTION

[0024] The present invention will be described in detail below with reference to the accompanying drawings in the embodiments of the present invention. However, the protection scope of the present invention is not limited to the following embodiments. That is, any simple equivalent changes and modifications made within the scope of the patent application of the present invention and the contents of the specification are still within the scope of the patent of the present invention.

[0025] like Figure 1-3 As shown in the figure, it includes the perception layer, network layer, platform layer and application layer. The perception layer deploys multiple types of sensors to collect real-time data on equipment operating parameters, dam structural status, reservoir hydrological environment and perimeter images, including underwater sonar, drones and cameras to achieve all-round monitoring;

[0026] The network layer adopts 5G+LoRa dynamic fusion networking, integrates the Deep Q network routing algorithm through the intelligent gateway, dynamically switches between 5G high-speed links and LoRa wide-area links based on link quality, and configures a satellite communication backup link;

[0027] The platform layer builds a digital twin-driven big data analysis platform, which uses AI algorithms such as LSTM and YOLOv5 to predict equipment failures and identify abnormal events, and achieves real-time bidirectional synchronization between virtual models and physical power plants.

[0028] The application layer supports hierarchical early warning, automatic triggering of emergency strategies, and three-dimensional visual remote operation and maintenance, and has multi-terminal access capabilities.

[0029] According to the above description, the network layer's Deep Q network routing algorithm monitors signal strength, latency, and packet loss rate in real time, giving priority to ensuring low-latency transmission of real-time data, while reducing power consumption of low-speed data transmission, and realizing intelligent path optimization of multi-network integration. The Deep Q network intelligent routing intelligent gateway collects link status parameters in real time and dynamically decides on the transmission path through the Deep Q network algorithm.

[0030] According to the above description, it is further explained that the digital twin platform uses CFD fluid mechanics and structural mechanics models to simulate extreme scenarios such as flood impact and equipment fatigue, rehearse the impact of failures and optimize emergency strategies, forming a closed-loop control of physical equipment and virtual models. The edge node pushes the device status to the digital twin platform every 200ms, driving the real-time update of the 1:1 three-dimensional model built by the Unity engine. The fluid simulation module synchronously simulates the water flow pressure distribution. The platform layer adjusts the generator power factor every 5 minutes based on the load forecast results and sends it to the PLC controller through the Modbus protocol. The instructions contain a 64-bit dynamic check code, and the edge node verifies the legitimacy of the instructions before execution.

[0031] According to the above description, the security protection system further elaborates that it includes a network layer firewall, data layer AES-256 encryption and application layer instruction secondary verification mechanism, combined with the AI ​​threat detection model to block abnormal attacks in real time and switch to the backup link. When the vibration amplitude of the equipment exceeds the threshold of 80%, the local sound and light alarm is activated, and an early warning is pushed to the operation and maintenance APP through the MQTT protocol. The operation and maintenance personnel must confirm and deal with it within 2 hours.

[0032] According to the above description, the following steps are further elaborated: S1, the sensor collects device / environment data in real time, and generates a standardized data frame through noise reduction and feature extraction;

[0033] S2, selects the optimal path through 5G+LoRa fusion link and deep Q network algorithm, and switches to satellite communication in case of abnormality to ensure reliable data upload;

[0034] S3: Input data drives the synchronous update of the virtual model, and the AI ​​algorithm integrates analysis to generate equipment failure probability and environmental risk assessment results;

[0035] S4. Trigger audible and visual alarms, remote notifications, and automatic equipment control based on risk levels. Instructions are verified twice by edge nodes to prevent attacks.

[0036] According to the above description, it is further explained that in the hierarchical linkage response, the red warning triggers a satellite SMS to notify downstream evacuation, and at the same time records the emergency process and feeds it back to the digital twin platform to achieve self-learning optimization of the emergency plan.

[0037] Working Principle: This unmanned intelligent safety control system and method for hydropower stations deploys vibration sensors (acquisition frequency ≥ 10kHz), temperature sensors (accuracy ±0.5°C), and pressure sensors on key equipment such as turbines, generators, and transformers. These sensors acquire over 30 operating parameters, including equipment vibration amplitude, bearing temperature, and lubricating oil pressure, in real time, building a "digital fingerprint" of equipment health. Displacement sensors and seepage sensors are embedded in the dam, and ultrasonic water level gauges and weather stations are deployed in the reservoir area. Combined with underwater sonar and drone inspections, these systems achieve three-dimensional monitoring of the dam's structural safety, reservoir hydrology, and surrounding environment. AI cameras are deployed around the perimeter, capturing video streams at 25 frames per second in real time. Frame extraction is performed by edge nodes, reducing cloud transmission pressure.

[0038] All sensor timestamps are synchronized using the Beidou timing module (±100ns accuracy). Clock offsets are checked every 10 minutes to ensure time alignment of multi-source data. Vibration data is monitored in real time based on the 3σ rule. Amplitudes exceeding three times the mean are marked as noise points. Three consecutive abnormal points trigger a sensor self-check process. FFT spectrum analysis is performed on the vibration signal to extract characteristic frequencies of bearing faults (e.g., 108Hz for the inner race and 82Hz for the outer race). The original time-domain signal is converted into a frequency-domain feature vector (with a compression ratio of 10:1), and feature data is output every two seconds.

[0039] Video key frames and device control commands are transmitted via 5G slice channels with a fixed 10Mbps bandwidth. End-to-end latency is monitored in real time, triggering a link quality warning when the latency exceeds 80ms three times in a row. Dam seepage and meteorological data are aggregated via a LoRa gateway, with a single gateway covering a 3km radius. Sensor power consumption is controlled below 10mA, and a gateway signal strength self-check is performed at 1:00 AM daily. The intelligent gateway collects real-time 5G signal RSRP (above -105dBm indicates a high-quality signal), LoRa signal-to-noise ratio (above 5dB indicates a usable signal), link latency (5G ≤ 50ms, LoRa ≤ 10s), and sensor battery life (low power mode is triggered if the battery falls below 20%) to construct a four-dimensional state vector. Simultaneously, based on a deep Q network algorithm (the decision model is updated every 10 seconds), the 5G link is prioritized for real-time data transmission, while the LoRa link is used for low-speed data transmission. When the signals of both networks fall below the threshold simultaneously, the system automatically switches to BeiDou short message mode, sending encoded key data every minute.

[0040] The edge node pushes device status to the digital twin platform every 200ms, driving real-time updates of the 1:1 3D model built by the Unity engine. The fluid simulation module simultaneously simulates water flow pressure distribution. The platform layer adjusts the generator power factor every 5 minutes based on load forecast results. The instructions are sent to the PLC controller via the Modbus protocol. The instructions contain a 64-bit dynamic check code, and the edge node verifies the legitimacy of the instructions before execution.

[0041] The LSTM model inputs 10-dimensional data, including equipment vibration, temperature, and load. Incremental model training is performed at midnight daily. When the predicted failure probability is ≥90%, a yellow alert is triggered, and a maintenance work order is generated simultaneously. The YOLOv5 model detects video key frames at 150 FPS, identifies perimeter intrusions and floating object accumulation, and uses an optical flow algorithm to determine motion trajectories and distinguish between normal inspections and illegal intrusions. Simultaneously, the random forest model integrates water level, rainfall, and dam displacement data to output hourly flood overtopping risk values ​​and simultaneously generate flood discharge scheduling recommendations.

[0042] When the vibration amplitude of the equipment exceeds the threshold of 80%, the local sound and light alarm is activated, and an early warning is pushed to the operation and maintenance APP through the MQTT protocol. The operation and maintenance personnel must confirm and deal with the situation within 2 hours. Orange warning (environmental level): The water level in the reservoir reaches 90% of the flood control limit water level. An email is automatically sent to the flood control headquarters, and the flood discharge gate preheating program is remotely started, and the flood control responsible persons in the downstream townships are notified simultaneously.

[0043] Red alert (power station level): Dam displacement change rate > 2mm / hour and seepage flow > 5m 3 / s, an "emergency shutdown of the entire plant" is immediately executed: ① The edge node sends a hard-wired shutdown command to the PLC, cutting off the generator excitation power within 100ms; ② The electric gate control system initiates a rapid shutdown procedure and simultaneously starts the backup diesel generator; ③ An evacuation notice is sent via Beidou short message to villages within 3km downstream of the reservoir area, repeated every 5 minutes until the signal is restored;

[0044] Before executing a control instruction, the edge node verifies the 64-bit dynamic token, the consistency between the current state of the device and the target state of the instruction, and the compliance of the preset logic. If any of the verifications fails, the instruction is blocked and a log is recorded. After each emergency response, the system automatically records the event data and enters it into the digital twin platform for virtual review at 2 a.m. every day. The PPO algorithm is used to optimize the warning threshold and control logic, and a strategy update report is generated every week.

[0045] It should be noted that the present invention is an intelligent safety control system and method for unmanned hydropower stations. The above-mentioned electrical components are all existing technology products. Those skilled in the art select, install and complete the circuit debugging work according to the needs of use to ensure that all electrical appliances can work normally. The components are all universal standard parts or components known to those skilled in the art. Their structure and principles can be known to those skilled in the art through technical manuals or through conventional experimental methods. The applicant does not make specific restrictions here.

[0046] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An unmanned intelligent safety control system and method for hydropower stations, comprising a perception layer, a network layer, a platform layer, and an application layer, characterized by: The perception layer deploys multiple types of sensors to collect real-time data on equipment operating parameters, dam structural status, reservoir hydrological environment, and perimeter images. This includes underwater sonar, drones, and cameras for all-round monitoring. The network layer adopts 5G+LoRa dynamic fusion networking, integrates the Deep Q network routing algorithm through the intelligent gateway, dynamically switches between 5G high-speed links and LoRa wide-area links based on link quality, and configures a satellite communication backup link; The platform layer builds a digital twin-driven big data analysis platform, which uses AI algorithms such as LSTM and YOLOv5 to predict equipment failures and identify abnormal events, and achieves real-time bidirectional synchronization between virtual models and physical power plants. The application layer supports hierarchical early warning, automatic triggering of emergency strategies, and three-dimensional visual remote operation and maintenance, and has multi-terminal access capabilities.

2. The unmanned intelligent safety control system and method for a hydropower station according to claim 1 is characterized by: The network layer's Deep Q network routing algorithm monitors signal strength, latency, and packet loss rate in real time, prioritizing low-latency transmission of real-time data while reducing power consumption of low-speed data transmission, thereby achieving intelligent path optimization for multi-network convergence.

3. The unmanned intelligent safety control system and method for a hydropower station according to claim 1 is characterized in that: The digital twin platform uses CFD fluid mechanics and structural mechanics models to simulate extreme scenarios such as flood impact and equipment fatigue, rehearse the impact of failures and optimize emergency strategies, forming a closed-loop control of physical equipment and virtual models.

4. The unmanned intelligent safety control system and method for a hydropower station according to claim 1 is characterized in that: The security protection system includes a network layer firewall, data layer AES-256 encryption and application layer instruction secondary verification mechanism, combined with an AI threat detection model to block abnormal attacks in real time and switch to backup links.

5. The unmanned intelligent safety control system and method for a hydropower station according to claims 1-4, characterized in that: The following steps are included: S1, sensors collect device / environment data in real time, and generate standardized data frames through noise reduction and feature extraction; S2, selects the optimal path through 5G+LoRa fusion link and deep Q network algorithm, and switches to satellite communication in case of abnormality to ensure reliable data upload; S3: Input data drives the synchronous update of the virtual model, and the AI ​​algorithm integrates analysis to generate equipment failure probability and environmental risk assessment results; S4. Trigger audible and visual alarms, remote notifications, and automatic equipment control based on risk levels. Instructions are verified twice by edge nodes to prevent attacks.

6. The unmanned intelligent safety control system and method for a hydropower station according to claim 5, characterized in that: In the hierarchical linkage response, the red alert triggers a satellite SMS notification for downstream evacuation, while recording the emergency process and feeding it back to the digital twin platform to achieve self-learning optimization of the emergency plan.

Images