Reservoir dam operation safety sky-ground work intelligent sensing system and operation method

Abstract

The invention relates to a reservoir dam operation safety sky-land project intelligent sensing system and an operation method, and relates to the technical field of hydraulic engineering safety monitoring. The system is composed of a sky-land water conservancy project integrated monitoring and sensing system, a self-adaptive sampling module, a layered distributed architecture and a software and hardware integrated module, and multi-source data such as deformation, seepage, stress strain, vibration and environmental quantity are cooperatively collected through five dimensions of sky domain, airspace, territory, water domain and work domain. The monitoring frequency is dynamically adjusted by using an adaptive sampling strategy, and data cleaning, standardization, space-time registration and fusion processing are completed through a distributed architecture to generate a high-quality comprehensive data set. The system can realize total-factor and whole-process refined monitoring, effectively eliminates data islands, improves data quality and monitoring efficiency, has high reliability, real-time performance and expandability, and provides powerful data support and decision basis for dam safety assessment and intelligent early warning.

Description

Technical Field

[0001] This application relates to the field of water conservancy project safety monitoring technology, and in particular to the "sky-ground-water engineering" intelligent sensing system and operation method for reservoir dam operation safety. Background Technology

[0002] Reservoirs and dams are crucial national water conservancy infrastructure projects, serving multiple functions including flood control, water supply, irrigation, power generation, and navigation. Their operational safety directly impacts the safety of life and property and the sustainable development of the social economy in downstream areas. With the increasing service life of dams and the frequency of extreme weather events, operational safety risks are constantly rising. Therefore, various safety monitoring systems have been established both domestically and internationally to promptly identify potential hazards and implement countermeasures through real-time data acquisition and analysis.

[0003] However, existing reservoir dam safety monitoring technologies still have the following common problems in long-term application: 1. Monitoring dimensions are limited, resulting in data silos: Current systems largely rely on point-type and contact-type sensors, such as manometers, piezometers, and inclinometers, focusing on the acquisition of local parameters and lacking comprehensive coverage and multi-dimensional sensing capabilities. Significant differences exist in data formats, acquisition frequencies, and transmission protocols among different monitoring devices and systems, resulting in isolated and fragmented data that hinders comprehensive cross-dimensional and cross-system analysis.

[0004] 2. The sampling strategy is rigid, resulting in insufficient timeliness and effectiveness of the data: Most monitoring systems use a fixed-period sampling mode, which cannot dynamically adjust the sampling frequency and resolution according to the dam's operating status, weather conditions, or emergencies. During periods of stable operation, a large amount of redundant data is generated, which not only consumes storage and transmission resources but also increases data processing costs; during critical periods such as flood season and earthquakes, sampling may be insufficient, resulting in the loss of key information.

[0005] 3. Low data integration, resulting in limited analysis and early warning capabilities: Existing systems lack sufficient depth in fusing multi-source heterogeneous data, often involving simple time series splicing or single-indicator comparison. They lack deep fusion algorithms based on statistics, signal processing, or machine learning, making it difficult to effectively uncover correlation features between different monitoring dimensions and limiting the accuracy of anomaly identification, trend prediction, and risk assessment.

[0006] 4. Insufficient system integration and environmental adaptability: Current monitoring systems are mostly composed of multiple vendors and technology platforms, lacking a unified hardware and software architecture and exhibiting poor compatibility between subsystems. In complex water conservancy engineering environments, affected by temperature and humidity changes, electromagnetic interference, and unstable communication, the systems are prone to data loss, delays, or failures, resulting in insufficient stability and scalability.

[0007] Based on the aforementioned issues, existing technologies cannot simultaneously achieve comprehensive monitoring, intelligent sampling, deep data fusion, and system integration and reliability. Therefore, there is an urgent need to develop a new type of reservoir dam operation safety monitoring system capable of integrating multi-source heterogeneous monitoring data, possessing adaptive sampling and control capabilities, efficient data fusion processing capabilities, and achieving deep software and hardware integration, in order to meet the practical needs of modern water conservancy project safety management. Summary of the Invention

[0008] The purpose of this application is to provide an intelligent sensing system and operation method for reservoir dam operation safety based on the "sky-ground-hydraulic engineering" framework, in order to solve the problems of data silos, rigid sampling strategies, low data fusion, and low hardware and software integration in reservoir dam safety monitoring systems in related technologies.

[0009] This invention provides an intelligent sensing system for reservoir dam operation safety, encompassing "sky-ground-water hydraulic engineering," which includes: The integrated monitoring and sensing system for dam construction includes a sky monitoring system, an airspace monitoring system, a ground monitoring system, a water area monitoring system, and an engineering area monitoring system. It is used to collaboratively collect multi-dimensional raw monitoring data on dam operation safety from five dimensions: sky, airspace, ground, water area, and engineering area, and output the raw monitoring data to downstream functional modules. An adaptive sampling module establishes a two-way data interaction relationship with the integrated monitoring and sensing system: on the one hand, it receives the raw monitoring data output by the integrated monitoring and sensing system; on the other hand, based on the fluctuation state of the raw monitoring data and the risk level of the environmental condition data, it dynamically generates sampling adjustment strategies (including sampling frequency and data resolution) for each monitoring system, and feeds the sampling adjustment strategies back to the integrated monitoring and sensing system, controlling the integrated monitoring and sensing system to collect and output target monitoring data according to the optimized sampling strategy. The hierarchical distributed architecture establishes a one-way data receiving relationship with the adaptive sampling module, which is used to receive the target monitoring data output by the adaptive sampling module. Through its built-in perception layer, transmission layer, data processing layer and application layer, the architecture sequentially realizes the collection and transmission, cleaning and standardization, spatiotemporal registration and fusion and safety assessment application of the target monitoring data, and finally generates the comprehensive analysis data required for dam safety early warning. The integrated hardware and software module establishes support relationships with the integrated monitoring and sensing system, the adaptive sampling module, and the hierarchical distributed architecture. It provides hardware adaptation and protection for the sensing terminals and data acquisition units of the integrated monitoring and sensing system; it provides embedded hardware for strategy calculation and algorithm execution software environment for the adaptive sampling module; and it provides data transmission network, fusion computing hardware, and security encryption software for the hierarchical distributed architecture, achieving hardware collaboration and software compatibility across the entire system.

[0010] In some embodiments, the integrated "sky-ground-hydraulic" monitoring and sensing system can be represented as: , in: The system employs BeiDou / GNSS reference stations, a spaceborne InSAR system, and high-resolution remote sensing satellites to monitor the overall displacement of the dam, surface deformation in the reservoir area, and regional subsidence. It includes UAV lidar, oblique photography system, infrared thermal imager, and multi-rotor inspection system, used to monitor dam surface deformation, crack development, temperature anomalies, structural defects, and leakage traces; The equipment includes surveying robots, total stations, 3D laser scanners, hydrostatic levels, video surveillance equipment, intelligent analysis systems, and environmental sensors, used to monitor dam displacement, tilt, settlement, and surrounding environmental parameters. These include piezometers, manometers, weirs, multibeam echo sounders, underwater robots, sonar equipment, ADCP current meters, water level gauges, and water temperature sensors, used to monitor seepage parameters, underwater structures, and hydrodynamic characteristics. The system employs upright plumb lines, inverted plumb lines, tension lines, inclinometers, strain gauges, distributed fiber optic temperature measurement systems, strong vibration meters, and accelerometers to monitor the internal deformation, stress, temperature field, and dynamic response of the dam body.

[0011] In some embodiments, the adaptive sampling module has a built-in multi-dimensional state evaluation unit and the following sampling modes: Tiered emergency sampling mode: When excessive flooding or rainfall exceeding the limit is detected, sampling is performed according to the formula. , Determine the risk level and increase the sampling frequency of all monitoring points to the highest level; High-frequency sampling mode: When the seismic intensity in the dam site area reaches a preset threshold, the sampling frequency of the strong-motion meter is adjusted according to the formula. Adjustment; Stable operation frequency reduction mode: During the stable operation phase, the sampling frequency is adjusted according to the formula. attenuation; Data fluctuation driven mode: When the monitored data exceeds the normal fluctuation range, the sampling interval of the relevant sensors is compressed. Specifically, when the measured data is greater than the abnormal threshold, the sampling interval of the relevant sensors is compressed; or when the real-time monitored data is greater than the preset abnormal threshold, the sampling interval of the relevant sensors is compressed. Abnormal threshold .

[0012] in: This indicates the warning threshold for inbound flow. K Indicates the risk coefficient; λ represents the total capacity of the reservoir; λ represents the safety factor. Indicates the dominant earthquake frequency; Indicates the sampling redundancy coefficient; Indicates the sampling frequency; The attenuation coefficient is represented by t; the steady-state operating time is represented by t. Indicates the historical mean; n represents the sensitivity coefficient; Indicates historical standard deviation; This indicates an abnormal threshold. Real-time monitoring data refers to the real-time acquisition data of a specific monitoring indicator, including real-time measured values ​​of specific monitoring parameters such as seepage pressure, displacement, strain, and water level. This is used to quantitatively determine whether the current measured data deviates from the historical stable range, thereby triggering an adjustment to the sampling strategy.

[0013] In some embodiments, the adaptive sampling module has a self-learning function, automatically correcting the sampling parameter threshold based on historical data and early warning feedback. Its update formula is as follows: , in: Indicates the updated threshold; Indicates the threshold before the update; Indicates the learning rate; This represents the sampling error of the i-th time. This represents the target error.

[0014] In some embodiments, the layered distributed architecture includes: Sensing layer: Integrates various sensors and data acquisition terminals in the "sky-ground-water" system; Transport layer: A hybrid network using fiber optics as the primary source and 5G or LPWAN as a secondary source, with a communication error rate not exceeding 10. -4 It has the function of switching between primary and backup dual channels; Data processing layer: Performs data cleaning, standardization, and spatiotemporal registration, and fuses five core datasets using the Kalman filter algorithm. Generate comprehensive data. ; Application layer: Responsible for visualizing monitoring results, generating reports, and issuing early warning information; in: This represents the deformation dataset, including monitoring results such as horizontal displacement, vertical settlement, and tilt angle of the dam body and foundation. This represents a seepage dataset, including parameters such as seepage pressure, seepage flow rate, pore water pressure, and water level. This represents a stress-strain dataset, including structural stress, strain distribution, and temperature-compensated stress-strain coupling data. This represents a vibration dataset, including ground motion, dam dynamic response, and spectral characteristic parameters. This represents a dataset of environmental quantities, including environmental influencing factors such as temperature, humidity, wind speed, rainfall, and external loads. This represents a multi-source data fusion function based on Kalman filtering, used to fuse five core datasets under spatiotemporally consistent coordinates and a time reference, to obtain comprehensive monitoring data with minimized noise and optimal accuracy. .

[0015] In some embodiments, the Kalman filtering algorithm of the data processing layer includes the following steps: Based on the state prediction value at the previous moment and measured values Calculate Kalman gain : , Updated state estimate: , Where H is the observation matrix, Let R be the covariance matrix, and let R be the measurement noise covariance matrix. For the estimation of the fused state, For measurement vectors, This is the Kalman gain.

[0016] In some embodiments, the integrated hardware and software module includes: Hardware components: industrial-grade embedded computing devices, GPU acceleration units, UPS uninterruptible power supplies, lightning protection and grounding systems, and industrial firewalls; Software components: A unified data acquisition and control platform, compatible with multi-protocol sensor access; a cloud-based analysis platform that integrates data fusion and risk assessment algorithms, and has access control and data encryption functions.

[0017] An operational method for a "sky-ground hydraulic engineering" intelligent sensing system for reservoir dam operation safety includes the following steps: I. Multi-dimensional raw data acquisition: Activate the integrated monitoring and sensing system of sky, ground and hydraulic engineering, and collect multi-dimensional raw monitoring data (including data on deformation, seepage, stress and strain, vibration and environmental quantities, etc.) of dam operation safety from five dimensions: sky, air, ground, water, and engineering, according to the preset initial sampling plan, and output the raw monitoring data to the adaptive sampling module in real time. II. Adaptive Sampling Dynamic Adjustment: The adaptive sampling module receives the raw monitoring data output from step I, analyzes the fluctuation state of the raw monitoring data and the risk level of the environmental operating conditions data through the built-in multi-dimensional state evaluation unit, and dynamically generates sampling adjustment strategies (including high-frequency sampling under risky operating conditions, low-frequency sampling during stable phases, and encrypted sampling when data is abnormal). The sampling adjustment strategies are fed back to the integrated monitoring and sensing system, which controls it to collect and output target monitoring data according to the optimized strategies. III. Reliable Transmission of Target Data: The hierarchical distributed architecture's transmission layer receives the target monitoring data output from step II, transmitting it through a hybrid communication network primarily based on fiber optics and supplemented by 5G / LPWAN (communication error rate ≤10⁻). 4 It has a dual-channel switching function (primary and backup), which transmits target monitoring data to the data processing layer in real time without loss; IV. Data Processing and Deep Fusion: The data processing layer receives the target monitoring data transmitted in step III, first performing data cleaning (removing outliers using the 3σ principle), format standardization, and spatiotemporal registration (based on a unified timestamp and CGCS2000 coordinate reference), and then using the Kalman filter algorithm on five core datasets: , The data is then fused to generate comprehensive analytical data for risk identification and security early warning. V. Comprehensive Data Application and Early Warning: The application layer receives the comprehensive analysis data output from step IV, uses the GIS 3D visualization module to dynamically display the monitoring results, automatically generate monitoring reports, and issue graded early warning information according to preset early warning rules (monitoring values ​​exceed the threshold and the duration meets the requirements), while recording abnormal event data. VI. Self-learning optimization feedback: The system collects the early warning results, abnormal event data and historical monitoring data from step V. Through the self-learning unit of the adaptive sampling module, it automatically corrects the sampling parameter thresholds, such as the inflow warning threshold and the earthquake intensity trigger threshold, according to the update formula. It also optimizes the early warning threshold model of step V and feeds the optimized parameters back to steps II and V to achieve continuous iterative upgrades of the monitoring strategy and early warning model.

[0018] In some embodiments, the sampling strategy in step II includes: when > When the time comes, a tiered emergency sampling mode is entered, in which... For inbound flow, The inbound flow warning threshold; when > When entering the high-frequency seismic sampling mode, among which The measured seismic intensity (or measured ground motion acceleration intensity) of the dam site area. The preset trigger threshold for the high-frequency seismic sampling mode (or the seismic intensity warning threshold for the dam site area). When the fluctuation of the monitored data is less than the set threshold, it enters a stable operation frequency reduction mode; When measured data > Or when the measured data is greater than When the preset amplitude is reached, the system enters the data fluctuation-driven mode. This indicates the threshold for anomaly detection.

[0019] In some embodiments, step II, when any sampling mode is triggered, further includes cross-subsystem collaborative sampling frequency allocation: for the i-th sensor located within the event's influence range, its sampling frequency is allocated according to... Perform automatic adjustment. in: This represents the new sampling frequency of the i-th sensor. Indicates the basic sampling frequency. This indicates the distance between the sensor and the event center. Indicates the maximum distance of the trigger range. Indicates the distance attenuation coefficient; Furthermore, the cross-subsystem sampling timestamp synchronization error does not exceed ±50ms.

[0020] The beneficial effects of the technical solution of this invention include: 1. Comprehensive and multi-dimensional monitoring, breaking down data silos. Through the five-dimensional monitoring system of "sky, ground, and water engineering", the system achieves full coverage of all elements, from macroscopic surface deformation, dam surface defects, internal stress and strain to seepage dynamics and underwater structures. All types of monitoring data are standardized in terms of time reference, spatial coordinates, and data format. The system also eliminates data isolation through multi-source fusion algorithms (Kalman filtering, spatiotemporal registration, etc.), forming a complete data foundation that can be directly used for multidimensional analysis, providing sufficient input conditions for subsequent risk identification and prediction.

[0021] 2. Intelligent and efficient sampling The adaptive sampling module combines real-time status quantity identification with multiple strategy modes (tiered emergency response, high-frequency sampling, low-frequency sampling, local encryption, etc.). During stable operation, it reduces the sampling frequency through an exponential decay model, reducing redundant data storage and transmission pressure by more than 50%. In the event of emergencies such as floods and earthquakes, it automatically switches to high-frequency sampling mode, increasing the sampling frequency of key parameters several times to ensure data integrity and timeliness when anomalies occur. This mechanism effectively balances system resource utilization and monitoring accuracy.

[0022] 3. Deep Data Integration The system introduces a deep fusion mechanism at the data processing layer, which not only performs noise filtering and interpolation correction on similar monitoring data, but also establishes a joint model through cross-dimensional multi-source correlation (such as GNSS and InSAR, seepage pressure and flow velocity, strain and temperature field, etc.), thereby improving the accuracy of anomaly feature extraction and trend prediction. The high-quality dataset generated after fusion provides a stable and reliable input for the risk assessment model, enhancing the scientific nature of analysis and early warning.

[0023] 4. The system is stable and reliable. The integrated hardware and software module adopts industrial-grade processing units, IP66-level protective enclosures, lightning protection grounding, and redundant UPS power supply to ensure long-term operational stability in complex environments such as high humidity, temperature differences, and electromagnetic interference. On the software side, it adopts a real-time operating system (RTOS) and a distributed computing framework to achieve parallelization of low-latency task response and large-scale historical data processing, ensuring that the system has high reliability, real-time performance, and scalability.

[0024] 5. Highly efficient closed-loop management The method forms a closed-loop operation mechanism of "sensing-sampling control-data processing-application and early warning" in terms of system architecture; each link interacts with control commands in real time through standardized interfaces, so that the monitoring strategy can be adjusted in time according to the analysis results, and the early warning signal can directly drive the switching of sampling strategy, forming a fast and closed-loop operation and management link, which significantly improves the initiative and response speed of dam operation and safety management. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 A schematic diagram of the overall system architecture provided in the embodiments of this application; Figure 2This application provides an integrated monitoring and sensing system architecture diagram for "sky-ground-hydraulic engineering" as shown in the embodiments. Figure 3 The system operation flowchart provided for the embodiments of this application; Figure 4 A flowchart of data fusion processing provided for embodiments of this application; Figure 5 An adaptive sampling control logic diagram provided for embodiments of this application; Figure 6 This is a schematic diagram of a layered distributed architecture provided in an embodiment of this application. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0028] This application provides an intelligent sensing system and operation method for reservoir dam operation safety based on the "sky-ground-water hydraulic engineering" framework. It can solve problems such as data silos, rigid sampling strategies, low data fusion, and low hardware and software integration in reservoir dam safety monitoring systems in related technologies.

[0029] Please see Figures 1-6 The integrated monitoring and sensing system for "sky-ground-hydraulic engineering" is described in detail below: The integrated "sky-ground-hydraulic engineering" monitoring and sensing system is the core of this invention, aiming to establish a multi-dimensional monitoring network covering the entire space and all elements of the dam and reservoir area. This system is divided into five subsystems according to the monitoring space and objects: Sky Domain (… ),airspace( ),area( ), water area ( ) and industrial domain ( ).

[0030] Let the overall monitoring and sensing system include S, which can be represented as:

[0031] in: (Sky Domain Monitoring System): Employing BeiDou / GNSS reference stations, spaceborne InSAR systems, etc., it monitors macroscopic parameters such as the overall displacement of the dam and the surface deformation of the reservoir area, with positioning accuracy reaching centimeter to millimeter level, and captures long-term slow deformation; (Airspace monitoring system): Includes UAV lidar, infrared thermal imager, etc., to acquire information such as surface deformation and crack development. The lidar three-dimensional point cloud accuracy reaches the centimeter level, and the infrared thermal imaging can locate potential leakage hazards. (Regional monitoring system): including measuring robots, environmental sensors, etc., to monitor surface displacement at the millimeter level and environmental parameters, and intelligently analyze video to identify abnormal behavior; (Water area monitoring system): including piezometers, multibeam echo sounders, etc., to monitor the seepage field, underwater structure and hydrodynamic characteristics, and to keep track of seepage pressure, reservoir bottom siltation and other conditions in real time; (Industrial Area Monitoring System): Employing equipment such as vertical lines and strain gauges, it monitors internal deformation, structural stress, temperature field, and dynamic response, directly reflecting the internal structural state.

[0032] Each subsystem is equipped with targeted sensing equipment, data acquisition methods, and a unified interface protocol, forming a monitoring data foundation that can be synchronized in time and space.

[0033] Tianyu Monitoring System ( The explanation is as follows: The monitoring objectives of the Tianyu Monitoring System are: to acquire information on the overall displacement of the dam, surface deformation, subsidence trends, and macro-geological changes in the reservoir area and surrounding regions.

[0034] Core equipment and parameters:

[0035] Data processing methods of the Tianyu monitoring system: The displacement was calculated using the raw GNSS observation data through least squares adjustment: , in, Let A be the displacement vector to be determined, P be the design matrix, P be the weight matrix, and L be the observation vector. The InSAR data is processed using time-series differencing to obtain the relative deformation, and coordinate reference unification and accuracy correction are performed using GNSS results.

[0036] Airspace monitoring system ( The explanation is as follows: The airspace monitoring system monitors the following: surface deformation of the dam body, crack development, abnormal temperature distribution, and signs of leakage.

[0037] Core equipment and parameters:

[0038] Data processing methods for airspace monitoring systems: 3D point cloud data is registered using iterative nearest point (ICP) registration: , in, and Let R be the coordinates of the corresponding points in the two scans, R be the rotation matrix, and t be the translation vector. Calculate the pixel temperature difference matrix from the infrared image. It is used to locate areas with abnormal temperatures.

[0039] Regional monitoring system ( The explanation is as follows: The monitoring objective of the regional monitoring system is to acquire real-time data on dam surface displacement, tilt, settlement, and changes in the surrounding environment.

[0040] Core equipment and parameters:

[0041] Data processing methods for regional monitoring systems: The displacement field of the dam surface was fitted using B-spline surface fitting: , in, , As basis functions, These are the coordinates of the control points. This method can smoothly represent irregular monitoring grid data and generate a continuous displacement field.

[0042] The basis functions in this invention refer to B-spline basis functions, which have explicit mathematical expressions. They are defined as a class of piecewise polynomial functions, typically constructed using the Cox-de Boor recursive formula, and widely used in engineering scenarios such as curve and surface fitting and smooth displacement field reconstruction. They possess good locality, differentiability, and numerical stability. Wherein: , Let represent the i-th k-th B-spline basis function.

[0043] Water monitoring system ( The explanation is as follows: The monitoring objectives of the water area monitoring system are: to monitor seepage field parameters, underwater structure safety status, and reservoir hydrodynamic characteristics.

[0044] Core equipment and parameters:

[0045] Data processing methods for water monitoring systems: The seepage flow calculation uses Darcy's law correction: , Where Q is the seepage flow rate, K is the permeability coefficient, and A is the cross-sectional area of ​​the seepage flow. Where L is the head difference and L is the seepage path length.

[0046] Industrial monitoring system ( The explanation is as follows: The monitoring objectives of the engineering domain monitoring system are: to monitor the internal deformation, structural stress and strain, temperature field and dynamic response of the dam body.

[0047] Core equipment and parameters:

[0048] Data processing methods for industrial monitoring systems: Internal stress calculation is based on the generalized Hooke's law: , in, For stress tensor, For strain tensor, This represents the tensor of the material's elastic constants.

[0049] Regarding the cross-dimensional data management and interface unification of this invention: all five subsystems achieve data standardization through a unified sensor access protocol, time synchronization mechanism, and coordinate benchmark. Before data enters the processing layer, format conversion and identification information unification are completed to support subsequent Kalman filtering, multi-source fusion, and risk identification.

[0050] The sampling module of this invention is described as follows: The adaptive sampling module is the core of the system's strategy scheduling. It is used to dynamically adjust the sampling frequency, sampling interval, data resolution, and synchronization method of each monitoring subsystem under different operating conditions, thereby ensuring the integrity and adaptability of monitoring data in both the time and spatial domains. This module consists of a state awareness unit, a strategy decision engine, a sampling execution unit, and a self-learning optimization unit. Through close coupling with the "sky, ground, and water" five-dimensional monitoring system, it realizes the automated allocation of multi-source monitoring tasks.

[0051] The adaptive sampling module monitors the target and its effective range. The goal of the adaptive sampling module is to dynamically adjust the sensor's sampling parameters based on changes in environmental conditions, structural state evolution, and monitoring data characteristics. Its scope of application covers: Time-domain modulation: changing the sampling frequency, sampling interval, and data packetization period; Spatial collaboration: When a local event is triggered, sensors in nearby areas are linked to synchronously and encryptedly sample. Resolution switching: Increase sensor resolution during high-risk phases and revert to standard mode during stable phases; Synchronization mechanism: Sampling timestamps are accurately synchronized across subsystems with an error not exceeding ±50 ms.

[0052] The adaptive sampling module consists of the following components:

[0053] The adaptive sampling module adapts to the core sampling strategy: Adaptive sampling strategies are divided into four main categories, each of which includes triggering conditions, adjustment methods, and calculation formulas.

[0054] (1) Tiered emergency sampling mode Applicable to extreme hydrological events such as torrential rain and floods. The trigger condition is the inflow rate. Exceeding the warning threshold The threshold calculation formula is: , in: k represents the risk coefficient (dimensionless, ranging from 0.5 to 1.5). This indicates the total reservoir capacity (m³). This indicates the safety amplification factor (1.0~2.0).

[0055] Once triggered, this mode will instruct all key monitoring points to increase the sampling frequency to the highest level, for example, increasing the piezometer from 1 time / minute to 1 time / second.

[0056] This applies to earthquake events occurring in and near the dam site. The triggering condition is the measured earthquake intensity. Greater than the preset threshold (e.g., 10 gal). The formula for calculating the sampling frequency of a strong-motion seismometer is: , in: Indicates the dominant frequency of the earthquake (Hz); This represents the sampling redundancy coefficient (0.1~0.3).

[0057] In this mode, displacement gauges, strain gauges, etc., simultaneously enter high-frequency sampling (e.g., ≥100 Hz) and gradually decrease after the event ends.

[0058] (3) Stable operation frequency reduction mode This method is suitable for situations where the operation is stable and the fluctuations in monitored data are within normal range. The sampling frequency attenuation formula is: , in: Indicates the initial sampling frequency; e is the base of the natural logarithm, which is 2.718; This represents the attenuation coefficient, with values ​​between 0.001 and 0.01. t represents the stable running time (in hours).

[0059] This mode can reduce data redundancy and transmission pressure.

[0060] (4) Data fluctuation driven mode This is triggered when real-time data shows abnormal fluctuations; the judgment condition formula is as follows: , in: Indicates the historical average; Indicates historical standard deviation; n represents the sensitivity coefficient (2~4). This indicates the abnormal threshold.

[0061] Real-time monitoring data refers to the real-time acquisition data of a specific monitoring indicator, including real-time measured values ​​of specific monitoring parameters such as seepage pressure, displacement, strain, and water level. This is used to quantitatively determine whether the current measured data deviates from the historical stable range, thereby triggering the adjustment of the sampling strategy. Specifically, when the real-time monitoring data is greater than the abnormal threshold, the sampling interval of the relevant sensor is compressed; or when the real-time monitoring data is greater than the abnormal threshold by a certain amount, the sampling interval of the relevant sensor is compressed.

[0062] Once triggered, the sampling interval at the measurement point is shortened, and nearby sensors enter a synchronous encryption mode to capture the event evolution process.

[0063] Adaptive sampling module learning optimization mechanism: The self-learning unit automatically adjusts various thresholds and strategy parameters based on historical data and early warning feedback. The update formula is: , in: Indicates the updated threshold; Indicates the threshold before the update; This represents the learning rate (0.01~0.1). This represents the sampling error of the i-th time. This represents the target error, and the target error value. The settings are based on typical operational data and historical early warning results, using empirical statistics and engineering backtracking analysis. For example, by statistically analyzing the mean and variance of sampling errors under normal operating conditions, an empirical tolerance threshold is set, or the backtracking average of system response deviations in previous false alarms / missed alarms is used as the target value. This applies to different monitoring objects (such as seepage pressure, displacement, strain, etc.). It can be set individually or dynamically adjusted according to the risk level to ensure that the model has good adaptability and robustness.

[0064] This mechanism can gradually reduce the probability of false alarms and missed alarms over a long period of operation.

[0065] Adaptive sampling module number configuration and execution logic: Example of parameter configuration:

[0066] The execution logic includes: state awareness → policy matching → instruction distribution → sampling adjustment → data feedback → state update. The entire process latency is controlled within 1 second.

[0067] Adaptive sampling module subsystem collaborative sampling: When a subsystem triggers a sampling adjustment, the adaptive sampling module will invoke a cross-system cooperative algorithm: Cooperative sampling frequency allocation formula: , in: This represents the new sampling frequency of the i-th sensor. Indicates the basic sampling frequency. This indicates the distance between the sensor and the event center. Indicates the maximum distance of the trigger range. This represents the distance attenuation coefficient (0.1~0.5). This mechanism enables the monitoring of a gradual distribution of density in space.

[0068] The layered distributed architecture of this invention is described below: The layered distributed architecture is the data organization and processing backbone of the system of this invention, responsible for the efficient transmission, processing, and application of heterogeneous monitoring data from the "sky, ground, and water" multi-dimensional sensing system. This architecture follows the principles of layered management, module decoupling, parallel processing, and redundancy assurance, and is divided into four layers from bottom to top: the sensing layer, the transmission layer, the data processing layer, and the application layer. Each layer has independent functions and interfaces, and interaction between upper and lower layers is achieved through standardized communication protocols.

[0069] The hierarchical distributed architecture design goal of this invention is as follows: Structured data flow: Different types of data flow sequentially according to the links of collection, transmission, processing, and application, avoiding cross-layer confusion.

[0070] Modular deployment: Each layer can be upgraded, replaced or expanded independently, reducing the difficulty of system maintenance.

[0071] Distributed processing: Supports multi-node parallel computing and load balancing, improving the ability to process large-scale data.

[0072] Redundancy and fault tolerance: Critical links and nodes have dual backup mechanisms to ensure continued operation in the event of a single point of failure.

[0073] The layered distributed architecture perception layer of this invention: The perception layer is responsible for raw data acquisition and is the bottom layer of the entire architecture. This layer integrates data from five monitoring subsystems ( , , , , ( ) sensors and data acquisition terminals.

[0074] Example of a perception layer device interface specification:

[0075] The data timestamps of the perception layer are uniformly controlled within ±1 ms, and the coordinate reference is uniformly set to the National Geodetic Coordinate System CGCS2000.

[0076] The layered distributed architecture of this invention includes the following transport layer: The transport layer's task is to transmit data collected by the sensing layer to the data processing layer with low latency and high reliability, supporting a hybrid wired and wireless network architecture. The main link uses a fiber optic ring network, while the secondary links use 5G NR or LPWAN (LoRa / NB-IoT) for redundancy.

[0077] Communication performance calculation formula: Link bandwidth requirement estimation: , in: N represents the number of sensors; This represents the sampling frequency (Hz) of the i-th sensor. Indicates the size of a single sampled data (bytes); This indicates the proportion of protocol overhead (0.05~0.3).

[0078] Delayed budget formula: , in: Indicates signal propagation delay; Indicates data transmission delay; This indicates a delay in protocol processing; This indicates a delay in queuing.

[0079] Link redundancy switching employs a primary / backup dual-channel system, with switching time controlled within 50ms, based on the link status detection formula: , in The packet loss rate is denoted by RTT (Round Trip Time), and the threshold is configured by the system.

[0080] The layered distributed architecture data processing layer of this invention: The data processing layer, located between the transport layer and the application layer, is responsible for data preprocessing, fusion, and structured storage. Its core functions include data cleaning, integrity verification, format standardization, spatiotemporal registration, data fusion, and result caching.

[0081] The data processing layer performs data cleaning, standardization, and spatiotemporal registration, and fuses five core datasets using the Kalman filter algorithm. To generate comprehensive data: , in: This represents the deformation dataset, including monitoring results such as horizontal displacement, vertical settlement, and tilt angle of the dam body and foundation. This represents a seepage dataset, including parameters such as seepage pressure, seepage flow rate, pore water pressure, and water level. This represents a stress-strain dataset, including structural stress, strain distribution, and temperature-compensated stress-strain coupling data. This represents a vibration dataset, including ground motion, dam dynamic response, and spectral characteristic parameters. This represents a dataset of environmental quantities, including environmental influencing factors such as temperature, humidity, wind speed, rainfall, and external loads. This represents a multi-source data fusion function based on Kalman filtering, used to fuse five core datasets under spatiotemporally consistent coordinates and a time reference, to obtain comprehensive monitoring data with minimized noise and optimal accuracy. .

[0082] Data preprocessing workflow: Cleaning: Outliers are removed using the 3σ principle, and missing values ​​are filled using linear interpolation / Kalman prediction.

[0083] Verification: CRC32 checksum comparison to ensure error-free data transmission.

[0084] Standardization: Converting to a unified physical unit and coordinate system.

[0085] Spatiotemporal registration: Multi-source alignment is achieved by using synchronized timestamps and spatial location identifiers.

[0086] Multi-source data fusion formula (Kalman filter): , in: This represents the state estimate after fusion; Represents the measurement vector; Indicates Kalman gain; H Represents the observation matrix; R represents the covariance matrix; R represents the measurement noise covariance.

[0087] H It is a linear mapping matrix connecting the dam state vector, the safety index to be monitored, and the observation vector, the measured data from the sensors. It can be expressed by the formula... Further refinement, among which For observation noise, state vector x Core indicators of dam safety, such as x =[ x Horizontal displacement, x Vertical displacement x osmotic pressure T ; Observation vector y Sensor measured data, for example y =[ y GNSS level, y The hydrostatic level is vertical. y [Pyrometer] T . H The elements directly correspond to the correlation strength between a sensor observation and a certain state index. For example, when GNSS only observes horizontal displacement, H The element corresponding to the horizontal displacement state is 1, and the element corresponding to the seepage pressure state is 0.

[0088] State covariance matrix P To describe the uncertainty of the dam's state estimates, the diagonal elements represent the variance of the estimation errors for each state index. For example, if the standard deviation of the initial displacement estimation error is set to 10 mm, then... P The corresponding element is 100mm. 2Off-diagonal elements represent uncertainties in the relationship between states, such as the correlation between displacement and seepage pressure, determined by engineering experience or historical data, and updated using the Kalman update formula. Dynamic iteration.

[0089] Measurement covariance matrix R This provides a statistical description of sensor observation errors. Elements are directly related to sensor accuracy parameters, and diagonal elements represent the error variance of a single sensor. For example, if the GNSS plane accuracy is ≤3mm, then... R The corresponding element is (3mm). 2 =9mm 2 If the piezometer accuracy is ±0.5% FS (range 0-1000kPa), then the error is ±5kPa. R The corresponding element is (5 kPa). 2 =25kPa 2 Off-diagonal elements represent the correlation of sensor errors. For example, if the observation errors of GNSS and piezometer are independent, then the off-diagonal elements are 0.

[0090] Example of a data classification storage structure:

[0091] The layered distributed architecture of this invention includes the application layer. The application layer provides end users with functions such as data visualization, report generation, and alert publishing, and interfaces with external management systems. The application layer provides data access interfaces via APIs, supporting access from web, mobile, and desktop devices.

[0092] Early warning issuance determination formula: , in: Indicates the monitored value; Indicates the warning threshold; Indicates the duration of exceeding the threshold; This indicates the minimum duration requirement.

[0093] The GIS 3D visualization module at the application layer is linked with the monitoring data in real time, enabling multi-timescale playback and prediction model invocation.

[0094] This invention relates to a layered distributed architecture and cross-layer data flow and security mechanism: To ensure the security and integrity of data flow from the perception layer to the application layer, the system is designed with multiple security mechanisms, including: Link encryption: End-to-end TLS 1.3 encryption at the transport layer.

[0095] Access control: Application layer permission management based on RBAC (Role-Based Access Control).

[0096] Data integrity verification: SHA-256 hash verification is performed during both the transmission and storage phases.

[0097] The integrated hardware and software module of this invention is described as follows: The integrated hardware and software module forms the foundation for the system's operation, providing end-to-end hardware support and software runtime environment from data acquisition, transmission, processing to application. This module emphasizes the collaborative optimization of hardware and software, achieving stable system operation, rapid response, and scalable deployment through the unified integration of embedded computing platforms, real-time operating systems, industrial-grade protection equipment, and cloud-based distributed platforms.

[0098] The design goals of the integrated hardware and software module are: Uniformity: Ensure the consistency of hardware interfaces, communication protocols, and software operating environments among the various monitoring subsystems; High reliability: Meets the requirements for long-term trouble-free operation in harsh environments; Scalability: Supports smooth expansion of hardware modules and software functions; Rapid response: The collected data is processed and put into application within the shortest possible time.

[0099] The integrated hardware and software module's hardware component includes a core processing unit, power supply system, protection and environmental adaptability, and network and security equipment. (1) Core processing unit Industrial-grade embedded computing devices, using ARM multi-core processors or x86 architecture CPUs with a main frequency of ≥ 2.0GHz, and supporting SIMD instruction sets to accelerate matrix operations; Configure GPUs (CUDA / OpenCL supported) to accelerate point cloud processing, image analysis, and deep learning inference; Memory capacity ≥ 16 GB, storage device is industrial grade SSD (≥ 1 TB) and supports NVMe interface.

[0100] (2) Power supply system The main power source is from the mains power or the hydroelectric power station's self-powered system, while the auxiliary power sources are uninterruptible power supplies (UPS) and solar panels; UPS battery life calculation formula: , in: Indicates battery capacity (Ah); Indicates battery voltage (V); This indicates the system conversion efficiency (0.85~0.95). This indicates the system load power (W).

[0101] (3) Protection and environmental adaptability The enclosure has an IP66 protection rating and a temperature range of -40 ℃ to +70 ℃. Lightning protection grounding resistance ≤ 4Ω; It is equipped with moisture-proof heating components and shock-absorbing devices.

[0102] (4) Network and security equipment Gigabit industrial switches that support VLAN partitioning and QoS priority management; Industrial firewalls support DPI (Deep Packet Inspection), Intrusion Detection System (IDS), and Intrusion Prevention System (IPS). The Hardware Security Module (HSM) is used for storing encryption keys and performing hardware-level encryption operations.

[0103] The integrated hardware and software module's software component includes a data acquisition and control platform, a data processing and fusion platform, and a fusion algorithm management platform. (1) Data acquisition and control platform Supports multi-protocol access (Modbus TCP / RTU, OPC UA, MQTT, HTTP / HTTPS, NTRIP); Visually configure data acquisition tasks and data flow paths; Real-time device status monitoring and remote firmware upgrade (FOTA).

[0104] (2) Data processing and fusion platform Local edge computing nodes run a real-time operating system (RTOS) to ensure millisecond-level response for high-priority tasks; The cloud platform performs batch and stream processing based on distributed computing frameworks (such as Apache Spark / Flink); Storage capacity calculation formula: , in: N represents the number of sensors; Indicates the sampling frequency (Hz); Indicates the size of a single sampled data (bytes); Indicates the data retention time (in seconds).

[0105] (3) Fusion Algorithm Management Platform Supports dynamic loading of algorithms such as Kalman filtering, particle filtering, Bayesian inference, and deep learning; Algorithm version management and canary release mechanisms allow for the gradual replacement of algorithm modules without affecting overall operation.

[0106] (4) Safety and fault tolerance mechanism User permissions are managed based on RBAC (Role-Based Access Control); System response time calculation formula: , in: Indicates the data collection time; Indicates data processing time; Indicates data transmission time; Indicates the data write time; Dual-machine hot standby switchover time ≤ 30 ms; Data is stored in multiple replicas, tolerating no data loss even if at least two nodes fail.

[0107] The hardware and software integration mechanism of the integrated hardware and software module includes load balancing strategies and collaborative optimization: (1) Load balancing strategy Task allocation across nodes follows the principle of optimizing resource utilization. The load balancing calculation formula is as follows: , in: This represents the task weight assigned to the i-th node; This represents the available computing power of the i-th node, which is measured by GFLOPS (1 billion floating-point operations per second). M represents the total number of nodes; This indicates the total workload.

[0108] (2) Collaborative optimization Edge computing nodes prioritize low-latency tasks (such as strong earthquake monitoring and emergency seepage pressure monitoring), while cloud nodes handle computationally intensive fusion analysis and historical data modeling. The software layer automatically detects hardware status (temperature, voltage, CPU utilization) and dynamically adjusts task allocation when abnormalities occur to avoid single-node overload.

[0109] Example table of technical parameters for integrated hardware and software modules:

[0110] The technical solution of the present invention will now be described in conjunction with engineering embodiments: I. The system deployment of this invention in an engineering example is as follows: 1. Integrated monitoring and sensing system for "sky, ground, and water engineering" To achieve multi-dimensional and comprehensive monitoring of the dam's operational status, five types of subsystems are deployed in the dam area and surrounding regions: Tianyu Monitoring: Three BeiDou / GNSS reference stations were deployed on the stable mountains on both sides of the river. Combined with satellite-borne InSAR data, the overall displacement of the dam and the surface subsidence of the reservoir area were obtained. The monitoring frequency was once a day.

[0111] Airspace monitoring: Deploy one fixed-wing UAV (equipped with LiDAR) to perform quarterly 3D modeling of the dam body, and two multi-rotor UAVs to conduct weekly inspections of key areas.

[0112] Regional monitoring: Eight surveying robots and four total stations were deployed to monitor the dam displacement on an hourly basis; at the same time, twenty cameras and five panoramic cameras were set up to realize real-time monitoring of the dam surface and abutments; in addition, three meteorological stations were set up to obtain environmental parameters such as temperature, humidity, wind speed and rainfall.

[0113] Water area monitoring: Fifty piezometers, twenty manometer tubes and eight measuring weirs are installed to achieve real-time seepage monitoring; two underwater robots and a multibeam echo sounder are deployed to conduct underwater structure detection every six months; three ADCPs and five water level gauges are installed to measure flow velocity, flow direction and water level changes in real time.

[0114] Monitoring of the dam site: Ten positive / inverted plumb line monitoring systems, thirty multi-point displacement gauges, eighty strain gauges, and sixty thermometers are set up along the key sections of the dam body, and eight strong seismographs are deployed to monitor the internal deformation, stress and strain distribution, temperature field changes and seismic response of the dam body in real time.

[0115] 2. Configuration of the adaptive sampling module in the engineering embodiment of the present invention: To achieve dynamic optimization and efficient utilization of monitoring data, the following sampling strategy is configured: Inbound Flow Warning: Set an inbound flow warning value = 60,000 m³ / s. When the measured flow rate exceeds this value, the system automatically increases the sampling frequency of all measuring points to a higher level.

[0116] Earthquake triggering: When the measured ground motion acceleration intensity in the dam site area exceeds 10 gal, the strong motion meter immediately enters the 100Hz high-frequency sampling mode.

[0117] Stable operation with reduced frequency: The piezometer initially samples at 6 times / hour during the stabilization period, and automatically decreases to about 3.8 times / hour after 24 hours of stable operation, reducing data redundancy.

[0118] Abnormal fluctuation response: When the measured seepage pressure at a certain measuring point exceeds μ+3σ (where μ is the historical mean and σ is the standard deviation), or when the measured seepage pressure at a certain measuring point exceeds μ+3σ by a certain margin, the sampling frequency of that measuring point will be automatically increased to 1 time / minute, and surrounding measuring points will be linked to enter the encrypted sampling mode.

[0119] 3. Layered distributed architecture and software / hardware integration To ensure data transmission reliability and processing efficiency, the system adopts the following architecture: Transport layer: Establish an optical fiber ring network (1000 Mbps) and equip it with a 5G DTU module as a backup link. The core switch adopts a dual-machine hot standby mode to ensure high communication reliability.

[0120] Data processing layer: Four edge computing servers are set up to preprocess and merge the data, and then transmit it synchronously to the cloud data center via a dedicated line.

[0121] Application layer: Deploy a unified monitoring data platform that supports 3D visualization, automatic generation of monitoring reports, and releases early warnings according to risk level, notifying management personnel in real time.

[0122] Hardware and software integration: Industrial-grade industrial control computers are selected, running a customized Linux system, and a UPS is configured to provide continuous power for 8 hours with a grounding resistance of less than 1Ω. Network and data security are ensured through an industrial firewall and security gateway.

[0123] II. System Operation and Performance Verification During routine operation: The system significantly reduces data redundancy through an adaptive sampling strategy, resulting in an overall data load reduction of approximately 60%. The platform can display various monitoring indicators and comprehensive safety status in real time.

[0124] During the flood season response phase: When the inflow reaches 55,000 m³ / s, the system automatically triggers the high-frequency sampling mode; by integrating seepage pressure and displacement data, an uplift pressure-displacement correlation dataset is obtained, which promptly issues risk warnings and provides decision-making basis for scheduling drainage.

[0125] Self-learning optimization phase: After one year of actual operation, the system continuously optimizes the rainfall-seepage coupling model and earthquake triggering threshold based on historical monitoring data, effectively reducing false alarms and missed alarms, and further improving the accuracy and adaptability of early warning.

[0126] III. Summary of Implementation Results Through the above deployment and operation, this system has achieved: The three-dimensional perception and fusion of multi-source heterogeneous data breaks through the isolation limitations of traditional monitoring; Intelligent adaptive sampling significantly reduces data transmission and storage pressure while improving response speed under critical operating conditions; Distributed processing and cloud-edge collaboration ensure high reliability and real-time performance in complex environments; Continuous self-learning and optimization enable dynamic evolution of the monitoring model and precise risk identification.

[0127] The system has demonstrated high efficiency and stability in practical applications, and has broad prospects for widespread application in ensuring the safe operation of reservoir dams.

[0128] The working principle of this application is as follows: The intelligent sensing system and operation method for reservoir dam operation safety proposed in this invention, encompassing "sky, ground, and hydraulic engineering," relies on multi-dimensional sensing, intelligent sampling, distributed processing, and integrated hardware and software to achieve fully automated operation from full-space information collection to safety early warning. Its working principle can be divided into four stages: data sensing → adaptive sampling and control → distributed data processing and fusion → intelligent application and early warning dissemination. Each stage is interconnected and operates in a closed loop.

[0129] 1. Data perception stage 1.1 Construction of Five-Dimensional Perception Network This method first deploys five monitoring subsystems: "Sky-Ground-Hydraulic Engineering". Tianyu Monitoring ( ): Large-scale surface deformation and geological environment changes are obtained through BeiDou / GNSS reference stations, InSAR satellite interferometry, and high-resolution remote sensing images.

[0130] Airspace monitoring ( ): Using UAV LiDAR, oblique photography, and infrared thermal imaging, information on apparent defects and temperature field anomalies in the dam body is obtained.

[0131] Regional monitoring ( ): The dam surface displacement, settlement, tilt, and surrounding environmental parameters are collected using measuring robots, 3D laser scanners, and hydrostatic levels.

[0132] Water monitoring ( ): Deploy piezometers, weirs, ADCP flowmeters, and underwater robots to acquire seepage field, underwater structure, and hydrodynamic information.

[0133] Industrial monitoring ( ): The internal stress, strain, temperature and dynamic response of the dam body are collected by positive / inverted plumb lines, strain gauges, distributed fiber optic thermometry and strong-motion seismographs.

[0134] 1.2 Multidimensional Synchronization and Standardization All sensors use a unified time synchronization (PPS / NTP) and coordinate reference (CGCS2000) to ensure that data from different sources can be accurately aligned in time and space, avoiding information silos.

[0135] 2. Adaptive Sampling and Control Stage This stage is controlled by the adaptive sampling module, which dynamically adjusts the sampling strategy based on the real-time status. 2.1 State Awareness: Monitoring data is used to calculate state variables (such as flow rate, acceleration, temperature field fluctuations, etc.) through fast algorithms to generate state identifiers. .

[0136] 2.2 Strategy Decision: Based on the state identifier, the strategy engine matches the sampling pattern: Rainstorms and floods → Tiered emergency response mode (increase sampling frequency at key monitoring points); Earthquake events → High-frequency sampling mode (≥100 Hz); Stable operation → Reduced frequency mode (exponential decay reduces sampling frequency); Abnormal fluctuations → Local encrypted sampling mode.

[0137] 2.3 Formula-driven: Sampling adjustments are calculated based on preset formulas, such as using formulas... and formula Etc., to ensure that adjustment decisions are quantifiable and verifiable.

[0138] 2.4 Cross-system collaboration: When a certain subsystem is triggered to encrypt sampling, the neighboring area and other related subsystems are synchronously adjusted according to the distance attenuation formula to form a spatially gradual sampling density.

[0139] 3. Distributed Data Processing and Fusion Stage This method employs a layered distributed architecture, distributing data processing tasks across different computing nodes to ensure both high efficiency and fault tolerance. 3.1 Transmission Layer: The fiber optic ring network is the main link, and 5G NR / LPWAN is the backup link. The health status of the link is determined in real time by a formula based on latency and packet loss rate thresholds, and switching occurs within 50ms in case of an anomaly.

[0140] 3.2 Data Processing Layer: Preprocessing: outlier removal (3σ principle), missing measurement value filling (Kalman prediction), physical units and coordinate system 1.

[0141] Spatiotemporal registration: Aligning multi-source data based on timestamps and coordinate identifiers.

[0142] Fusion computing: The Kalman filter formula fuses data from different sources, reducing measurement noise and enhancing trend recognition capabilities.

[0143] 3.3 Storage and Indexing: Structured storage (HDF5, CSV, MiniSEED, etc.), and the establishment of temporal and spatial indexes to support fast retrieval and multi-scale analysis.

[0144] 4. Intelligent Application and Early Warning Issuance Phase 4.1 Real-time visualization: The application layer GIS 3D platform dynamically renders monitoring data, supporting playback at multiple time scales and calling prediction models.

[0145] 4.2 Threshold Determination and Early Warning: Formula for determining continuous over-limit: , in: Indicates the monitored value; Indicates the warning threshold; Indicates the duration of exceeding the threshold; This indicates the minimum duration requirement.

[0146] Once the threshold is triggered, the system issues warning information through multiple channels (SMS, APP push, scheduling platform API).

[0147] 4.3 Decision support: Push risk analysis results to the dispatch center, including trend prediction curves and estimates of the scope of event impact.

[0148] 5. Hardware and software collaborative protection Hardware layer: Industrial-grade embedded computing unit, GPU accelerator card, UPS redundant power supply, protective shell, and industrial firewall ensure long-term stable operation of the system in harsh environments.

[0149] Software layer: Real-time operating system (RTOS) runs low-latency tasks on edge nodes, while cloud-based distributed frameworks (Spark / Flink) are responsible for batch data processing and historical modeling.

[0150] Collaboration mechanism: The task load balancing formula allocates computing tasks according to the computing power of each node to prevent single node overload, and can dynamically migrate tasks when hardware temperature or voltage is abnormal.

[0151] 6. Closed-loop operation characteristics The various stages of the method of this invention form a closed loop: Data sensing provides input for sampling and control; Sampling control determines data quality and spatiotemporal distribution; Distributed processing integrates multi-dimensional data to generate usable information; Applications and early warning systems transform information into decision-making criteria, and in turn adjust sampling strategies accordingly.

[0152] This closed-loop mechanism ensures that the system is always in a dynamically optimal monitoring configuration under changes in environmental and structural conditions, thereby improving the real-time performance and accuracy of dam operation safety management.

[0153] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A reservoir dam operation safety sky-ground-water intelligent sensing system, characterized in that, It comprises: The integrated monitoring and sensing system of sky, land, water and engineering, which comprises a sky monitoring system, an airspace monitoring system, a regional monitoring system, a water area monitoring system and an engineering monitoring system, is used for collecting multi-dimensional original monitoring data of dam operation safety from five dimensions of sky, airspace, region, water area and engineering field; The adaptive sampling module receives the original monitoring data output by the integrated monitoring and sensing system, dynamically generates a sampling adjustment strategy for each monitoring system according to the fluctuation state of the original monitoring data and the risk level of the environmental working condition data, and feeds back the sampling adjustment strategy to the integrated monitoring and sensing system to control the integrated monitoring and sensing system to collect and output target monitoring data according to the optimized sampling strategy; The hierarchical distributed architecture is used for receiving the target monitoring data output by the adaptive sampling module, and through the built-in sensing layer, transmission layer, data processing layer and application layer, it realizes the collection and transmission of target monitoring data, cleaning and standardization, spatio-temporal registration fusion and safety evaluation application in turn, and finally generates comprehensive analysis data required for dam safety warning; The hardware and software integrated module provides hardware adaptation and protection for the sensing terminal and data acquisition unit of the integrated monitoring and sensing system, provides embedded hardware and algorithm running software environment for the adaptive sampling module, and provides data transmission network, fusion computing hardware and security encryption software for the hierarchical distributed architecture, realizing hardware cooperation and software compatibility of the whole system.

2. The reservoir dam operation safety sky-ground-water-engineering intelligent sensing system of claim 1, wherein: The "sky-ground-water-engineering" integrated monitoring and sensing system is represented as: ， Wherein: : Using Beidou / GNSS reference station, space-borne InSAR system, high-resolution remote sensing satellite, for monitoring overall displacement of dam, surface deformation in reservoir area and regional subsidence; : Including unmanned aerial vehicle laser radar, tilt photography system, infrared thermal imager, multi-rotor inspection system, used for monitoring dam surface deformation, crack development, temperature anomaly and structure apparent defect, leakage trace; : including measurement robots, total station, three-dimensional laser scanner, static level, video monitoring equipment, intelligent analysis system and environmental quantity sensors, used for monitoring dam displacement, inclination, settlement and surrounding environmental quantity parameters; : includes osmometer, piezometer, weir, multibeam bathymetric system, underwater robot, sonar equipment, ADCP current meter, water level meter, water temperature sensor, for monitoring seepage parameters, underwater structures and hydrodynamic characteristics; : Adopting normal vertical line, inverse vertical line, tension line, inclinometer, strain gauge, distributed optical fiber temperature measurement system, strong motion instrument, acceleration sensor, for monitoring internal deformation of dam body, stress, temperature field and dynamic response.

3. The reservoir dam operation safety sky-ground-water-engineering intelligent sensing system of claim 1, wherein: The adaptive sampling module has a multi-dimensional state evaluation unit and the following sampling modes: Hierarchical emergency sampling mode: when excessive flood or over-limit rainfall is monitored, the risk level is judged according to the formula ， And the sampling frequency of all measuring points is raised to the highest level; High-frequency sampling mode: when the seismic intensity at the dam site reaches the preset threshold, the sampling frequency of the strong motion instrument is adjusted according to the formula adjustment; Stable operation frequency reduction mode: in the stable operation stage, the sampling frequency is adjusted according to the formula attenuation; Data fluctuation driven mode: when the real-time data meet At this time, the trigger-related sensor sampling interval compression is triggered; wherein: represents the storage flow alert threshold value; K represents the risk coefficient; represents the total capacity of the reservoir; λ denotes a safety amplification factor; denotes a seismic dominant frequency; denotes a sampling redundancy coefficient; denotes a sampling frequency; denotes an attenuation coefficient; t denotes a stable operation time; denotes a historical mean value; n denotes a sensitivity coefficient; denotes a historical standard deviation; denotes an abnormal increment.

4. The reservoir dam operation safety sky-ground-water-engineering intelligent sensing system of claim 1, wherein: The adaptive sampling module has a self-learning function, which automatically corrects the sampling parameter threshold according to historical data and warning feedback, and its update formula is: ， wherein: denotes the updated threshold value; denotes the updated threshold value; denotes the learning rate; denotes the i-th sampled error; denotes the target error.

5. The reservoir dam operation safety sky-ground-water-engineering intelligent sensing system of claim 1, wherein: The hierarchical distributed architecture comprises: Sensing layer: integrating various sensors and acquisition terminals in the "sky-ground-water-engineering" system; Transmission layer: hybrid network mainly using optical fiber supplemented by 5G or LPWAN, communication error rate not greater than 10 -4 , with main and standby dual channel switching function; Data processing layer: performing data cleaning, standardization, spatio-temporal registration, and fusing five types of core data sets through Kalman filtering algorithm Generate comprehensive data. ； Application layer: responsible for monitoring result visualization, report generation and warning information release; wherein: denotes a deformation dataset; denotes a seepage dataset; denotes a stress-strain dataset; denotes a vibration dataset; denotes an environmental quantity dataset; denotes a multi-source data fusion function based on Kalman filtering, used for fusing the five types of datasets in the spatiotemporally consistent coordinates and the time reference to obtain comprehensive monitoring data .

6. The reservoir dam operation safety sky-ground-hydraulic intelligent sensing system according to claim 5, characterized in that: the Kalman filtering algorithm of the data processing layer comprises the following steps: Based on the state prediction value of the previous time and the measurement value , the Kalman gain is calculated : , updating the state estimation: , where H is an observation matrix, is a covariance matrix, R is a measurement noise covariance matrix, is a fused state estimate, is a measurement vector, is a Kalman gain.

7. The reservoir dam operation safety sky-ground-hydraulic intelligent sensing system according to claim 1, characterized in that: the hardware-software integrated module comprises: hardware part: industrial-grade embedded computing device, GPU acceleration unit, UPS uninterruptible power supply, lightning protection grounding system and industrial firewall; software part: unified data acquisition and control platform, compatible with multiple protocol sensor access; cloud computing-based analysis platform, integrating data fusion and risk assessment algorithms, and having permission control and data encryption functions.

8. A method for operating the reservoir dam operation safety sky-ground-hydraulic intelligent sensing system according to any one of claims 1 to 7, characterized in that: comprising the following steps: I. Data acquisition: start the sky-ground-hydraulic integrated monitoring and sensing system, and collect multi-dimensional raw monitoring data according to the preset initial scheme; II. Adaptive sampling control: the adaptive sampling module analyzes the environmental state and monitoring data in real time, and dynamically adjusts the sampling strategy according to the analysis results, including increasing the sampling frequency under risk conditions, reducing the sampling frequency in the stable operation stage, and compressing the sampling interval when data is abnormal; III. Data transmission: the transmission layer transmits the data adjusted by the adaptive sampling to the data processing layer in real time through the hybrid communication network; IV. Data processing and fusion: the data processing layer pre-processes the received data and fuses them using a data fusion algorithm to obtain comprehensive data for risk identification and early warning; V. Data application: the application layer visualizes the processed comprehensive data and analysis results, generates reports, and publishes warning information according to preset rules; VI. Self-learning and optimization: the system continuously optimizes the sampling parameter threshold and warning threshold model through self-learning ability based on historical data and abnormal event records, improving the adaptability of the system and the accuracy of data collection.

9. The method for operating the reservoir dam operation safety sky-ground-hydraulic intelligent sensing system according to claim 8, characterized in that: the sampling strategy in step II comprises: When > the hierarchical emergency sampling mode is entered, wherein is the warehouse inflow, is the warehouse inflow warning threshold; When > enters a high-frequency sampling mode of earthquake, wherein is the measured seismic intensity in the dam site area, is a preset triggering threshold of the high-frequency sampling mode of earthquake. when the monitoring data fluctuation is less than the set threshold, enter the stable operation and frequency reduction mode; When > threshold value, enters a data fluctuation driving mode, represents an abnormal increment.

10. The method for operating the reservoir dam operation safety sky-ground-hydraulic intelligent sensing system according to claim 9, characterized in that: when any sampling mode is triggered, step II further comprises cross-subsystem coordinated sampling frequency allocation: for the i-th sensor located within the event influence range, the sampling frequency is allocated according to Perform automatic adjustment. wherein: denotes the new sampling frequency of the i-th sensor, denotes the base sampling frequency, denotes the distance of the sensor to the event center, denotes the maximum distance of the trigger range, denotes the distance decay coefficient; and the cross-subsystem sampling timestamp synchronization error is not more than ± 50 ms.

Images