Hydropower station dam safety monitoring data acquisition and transmission system

Abstract

The invention, which relates to the technical field of hydropower station dam safety monitoring, discloses a hydropower station dam safety monitoring data acquisition and transmission system comprising a cloud twin brain module and edge neurons. The cloud twin brain module comprises a sequence neural network engine and a reflection kernel generation module, the sequence neural network engine adopts a neural network architecture with parallel and cyclic dual representation, comprises a time mixing module and a channel mixing module, and can learn a normal operation mode of the dam from historical monitoring data; and the reflection nuclear generation module compresses the reference twin model into a lightweight reflection nuclear model and issues the lightweight reflection nuclear model to the edge device. The edge neuron comprises a micro-twinborn prediction module and a hierarchical transmission control module, and the micro-twinborn prediction module predicts a theoretical expected value of a dam state in real time and calculates a reflection deviation with an actual observation value; the hierarchical transmission control module implements a three-level response strategy according to the magnitude of the reflection deviation, transmits abstract information according to an abnormal trend, and uploads an emergency abnormality in time.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of dam safety monitoring of hydropower stations, and particularly relates to a dam safety monitoring data acquisition and transmission system for hydropower stations. BACKGROUND

[0002] As important water conservancy infrastructure, the safe operation of hydropower dams is related to the safety of people's lives and property and social and economic development downstream. With the rapid development of China's hydropower industry, dam safety monitoring technology has become a key technical means to ensure the long-term safe operation of dams. Traditional dam safety monitoring mainly relies on manual inspection and simple instrument measurement, which is difficult to meet the needs of modern dam safety management.

[0003] In recent years, with the development of Internet of Things, sensor technology and data communication technology, automatic monitoring systems have gradually become the mainstream technology for dam safety monitoring. Existing dam safety monitoring systems usually use distributed sensor networks to collect deformation, seepage, stress and other physical quantity data of various parts of the dam, and transmit the data to the monitoring center through wired or wireless communication networks for centralized analysis and processing. To some extent, this kind of system improves the automation level of monitoring and the real-time nature of data acquisition.

[0004] The existing technology has certain limitations through the traditional data acquisition and transmission method, such as: first, the traditional monitoring system uses a timing full-data transmission strategy, which requires continuous transmission of a large amount of monitoring data regardless of the normal operation state of the dam, resulting in serious waste of network bandwidth resources and increasing the system operation cost; second, the existing system mainly relies on simple threshold comparison for abnormal judgment, lacks deep understanding of the dam operation state, and cannot effectively distinguish between normal changes caused by environmental factors and real abnormal states, resulting in high false positive rate and false negative rate, affecting the reliability of the monitoring system; third, the traditional monitoring system lacks self-learning and evolution ability, and cannot optimize the monitoring strategy and improve the detection accuracy according to the accumulation of dam operation experience, with poor adaptability; fourth, the data processing of the existing system is mainly concentrated in the monitoring center, and the edge device only has simple data acquisition function and cannot perform local intelligent analysis, so the system reliability decreases when the network connection is unstable; fifth, the abnormal detection algorithm of the traditional system is relatively simple, and it is difficult to capture the complex change pattern of the dam operation state, especially for the recognition ability of gradual change and multi-variable coupling.

[0005] Therefore, there is an urgent need for a dam safety monitoring data acquisition and transmission system for hydropower stations with intelligent abnormal detection capability, on-demand data transmission, and self-learning evolution function to improve monitoring accuracy and efficiency, reduce system operation cost, and enhance the reliability and adaptability of the system. SUMMARY

[0006] The water power station dam safety monitoring data acquisition and transmission system aims to overcome the defects of the prior art and solve the above problems.

[0007] The water power station dam safety monitoring data acquisition and transmission system aims to overcome the defects of the prior art and solve the above problems. The twin brain module is deployed on a cloud server and includes: The sequence neural network engine adopts a neural network architecture with parallel and cyclic dual equivalent representation, and the architecture includes a time mixing module and a channel mixing module. The time mixing module performs weighted processing on dam multi-source historical monitoring data through a learnable time decay vector. The channel mixing module is used for modeling the coupling relationship across variables. The sequence neural network engine processes deformation data, seepage data, stress and strain data, reservoir water level data, rainfall data and environmental temperature data of the dam in parallel mode during training to generate a baseline twin model capable of representing the normal operating state of the dam. The reflection core generation module is used for quantization compression and parameter optimization of the trained baseline twin model, extraction of core inference logic, generation of a lightweight reflection core model with significantly reduced parameter scale, and distribution to edge neurons through network communication. And at least one edge neuron deployed on the dam monitoring site includes: The model and data interface module is used for receiving and loading the reflection core model from the cloud, establishing data connection with the on-site sensors, and real-time acquisition of physical quantity actual observation values and corresponding environmental variable observation values measured by the sensors. The data preprocessing module includes an instance normalization unit, a sequence blocking unit and a token generation unit, which is used for standardizing continuous observation value sequences, blocking according to a preset length and step, and converting the blocked data into block sequence tokens conforming to the input format of the reflection core model. The online twin prediction module calls the cyclic inference mode of the reflection core model, and based on the environmental variable observation values and historical state information at the current time, the cyclic inference mode with constant time complexity is used to real-time predict the theoretical normal value of each monitoring physical quantity under the current working condition, and the theoretical normal value sequence is used to form a dynamic reference trajectory. The reflection deviation calculation module calculates the Euclidean distance between the physical quantity actual observation value and the theoretical normal value as the reflection deviation value, and includes a deviation statistical analysis unit for calculating the statistical characteristics of the reflection deviation value in a sliding time window, including mean, variance, change trend coefficient and cumulative characteristics. The hierarchical transmission control module executes a three-level adaptive data transmission strategy according to the amplitude, duration and change rate of the reflection deviation value: when the reflection deviation value is below a first preset threshold, a first-level local record response is triggered; when the reflection deviation value is continuously higher than the first preset threshold but lower than a second preset threshold, a second-level summary information transmission response is triggered; when the reflection deviation value is instantaneously higher than the second preset threshold or the change rate exceeds a preset limit, a third-level immediate early warning information transmission response is triggered. The local storage module is used for caching observation data, reflection deviation values and system operation logs. The system further comprises a closed-loop model evolution mechanism, which receives uploaded summary information and early warning information data through the incremental learning module, performs incremental training and parameter fine-tuning on the benchmark twin model, and maintains iteration records through the model version management module. After the model is updated, the generation and distribution of a new version of the reflection kernel model are automatically triggered, realizing the continuous self-evolution ability of the system.

[0008] The time mixing module of the sequence neural network engine assigns different decay weights to data at different time steps in the historical monitoring data sequence, so that the influence weight of the historical data farther away from the current time is smaller, thereby effectively capturing long-term dependencies. The channel mixing module realizes the modeling of the correlation between different monitoring variables by performing information interaction and fusion in the feature dimension. The sequence neural network engine adopts a parallel computing mode in the training stage to improve training efficiency, and switches to a loop computing mode in the inference stage to realize efficient real-time inference.

[0009] The workflow of the reflection kernel generation module includes: model pruning and parameter quantization of the trained benchmark twin model to remove redundant parameters and calculation paths; extracting the core inference logic and key parameters of the benchmark twin model to generate a lightweight reflection kernel model suitable for edge computing environment; performing edge device compatibility testing on the generated reflection kernel model to ensure its stable operation in resource-constrained environment; and securely distributing the reflection kernel model to each edge neuron through encrypted transmission.

[0010] The prediction mechanism of the online twin prediction module includes: receiving the current time's reservoir water level observation value, environmental temperature observation value and rainfall observation value as input conditions; based on the environmental variable observation value and historical state information, calculating the expected value of each monitoring physical quantity at the next time through the loop inference calculation of the reflection kernel model; the sequence of continuously predicted expected values forms a dynamic reference trajectory, which reflects the normal behavior pattern of the dam under the current environmental conditions; the prediction process has constant time complexity, ensuring real-time response requirements.

[0011] The reflection deviation calculation module quantifies the degree of system deviation by calculating the Euclidean distance between the actual observation vector and the theoretical normal value vector; the deviation statistical analysis unit continuously tracks the changes in the reflection deviation value within a preset sliding time window, calculates its statistical characteristics including the average deviation, deviation variance, change trend coefficient and cumulative deviation amount, and is used to evaluate the stability and abnormal development trend of the system state.

[0012] The three-level response strategy of the hierarchical transmission control module is as follows: the first level is a local recording response, when the reflection deviation value is lower than the first preset threshold, the reflection deviation value and the corresponding observation value are stored in the local storage module, and no data transmission is performed to save communication resources; the second level is an abstract transmission response, when the reflection deviation value continuously exceeds the first preset threshold but is lower than the second preset threshold, a data abstract containing the deviation statistical characteristics, change trend analysis and time marker is generated, and network transmission is performed in a low priority mode; the third level is an early warning transmission response, when the reflection deviation value exceeds the second preset threshold or the deviation change rate exceeds the preset safety limit, the complete data record of the abnormal period is immediately intercepted, and real-time transmission is performed at the highest priority.

[0013] The instance normalization unit of the data preprocessing module independently normalizes the observation value sequence of each sensor channel, eliminating the differences in dimension and numerical range between different sensors; the sequence blocking unit splits the long time sequence according to fixed block length and overlap step, generating short sequence blocks suitable for reflection kernel model processing; the token generation unit converts each sequence block into a standardized input token containing position encoding and feature encoding.

[0014] The closed-loop model evolution mechanism realizes continuous learning through the following steps: the hierarchical transmission control module uploads the data generated by the second and third levels of response to the twin brain module; the incremental learning module performs quality evaluation and label annotation on the collected new data, distinguishing between normal changes, abnormal events and device failures; an incremental training algorithm is used to update the parameters of the baseline twin model, maintaining the memory of historical knowledge while learning new behavior patterns; the model version management module records the performance indicators and changes of each update, and triggers version release when the performance of the new model meets the preset improvement standard.

[0015] The incremental learning module uses an experience replay strategy to mix the newly collected data with historical representative data for training according to a preset ratio, preventing overfitting of the model to new data; a contrastive learning method is used to enhance the model's ability to recognize abnormal patterns, improving the model's discrimination performance by constructing positive and negative sample pairs; a multi-level model performance evaluation system is set up, including prediction accuracy indicators, abnormal detection rate indicators and false alarm rate indicators, to ensure the quality and reliability of model updates.

[0016] The hardware architecture of the edge neuron includes: an embedded computing unit, employing a low-power, high-performance processor and pre-set memory capacity, for running the reflection kernel model and data processing tasks; a multi-channel data acquisition unit, supporting simultaneous access of analog, digital, and pulse signals; a local storage unit, providing a pre-set data cache space to support local data backup and historical retrieval; a communication unit, supporting wireless and wired network communication, with communication redundancy and fault switching capabilities; and a clock synchronization unit, achieving high-precision time synchronization through a satellite positioning system to ensure the time consistency of distributed monitoring data.

[0017] The beneficial effects of this invention are: By constructing a cloud-based twin brain module, deep intelligent modeling of the dam's operational status was achieved. This twin brain module employs a sequence neural network engine with both parallel and recurrent representation capabilities, enabling efficient processing of massive historical monitoring data. During the training phase, it fully utilizes cloud-based parallel computing resources, significantly improving model training efficiency and convergence speed. Compared to traditional single-architecture neural networks, this invention's dual representation mechanism avoids the conflict between training efficiency and inference efficiency, achieving unified optimization of both.

[0018] The limbic neuron employs a micro-twin prediction mechanism to generate theoretical expected values ​​of the dam's state in real time on-site, achieving intelligent anomaly detection by calculating reflection deviations. This design fundamentally changes the traditional monitoring system's simple threshold alarm operation mode, enabling the system to possess true intelligent judgment capabilities. The limbic neuron can autonomously distinguish between normal changes caused by environmental factors and genuine abnormal states, significantly reducing false alarms and missed alarms.

[0019] The hierarchical transmission control mechanism dynamically adjusts the data transmission strategy based on the magnitude and trend of reflection deviation, achieving intelligent management of network resources. When the dam is operating normally, the system generates almost no data transmission demand, significantly reducing network bandwidth usage and communication costs. When a potential anomaly is detected, the system can automatically extract key information and compress it for transmission. In the event of an emergency anomaly, the system immediately triggers high-priority transmission to ensure the timely delivery of critical information. This intelligent transmission strategy guarantees both real-time monitoring and efficient utilization of communication resources.

[0020] The closed-loop model evolution mechanism enables the system to continuously learn and optimize itself. Through the incremental learning module, newly collected data is intelligently analyzed and the model is updated, allowing the system to continuously adapt to changes in the dam's operational status, learn new anomaly patterns, and improve detection accuracy. The introduction of experience replay strategies and comparative learning methods effectively prevents catastrophic forgetting, ensuring the system retains its historical experience while learning new knowledge. A multi-level model performance evaluation system provides reliable quality assurance for model updates, ensuring the directionality and stability of system evolution.

[0021] The reflection kernel generation and distribution mechanism enables efficient transfer of cloud-based intelligence to edge devices. Through model compression and parameter quantization, complex cloud models are converted into lightweight versions suitable for edge device operation, giving edge devices independent intelligent analysis capabilities. The incremental update mechanism only transmits the changed parts of the model parameters, significantly reducing the network transmission volume for model distribution and improving the efficiency and real-time performance of system updates.

[0022] An intelligent, hierarchical transmission strategy enables on-demand data transmission, minimizing network resource consumption while ensuring monitoring quality. Addressing the high false alarm rate and lack of intelligent judgment in existing systems, this invention utilizes micro-twin prediction and reflection deviation calculation to give the system a deep understanding of dam state changes, accurately distinguishing between normal and abnormal states. To address the limitations of traditional systems in terms of self-evolution and poor adaptability, a complete closed-loop learning mechanism is constructed, enabling the system to continuously optimize performance as operational experience accumulates.

[0023] The local intelligent analysis capabilities of edge neurons enable the system to function normally even under unstable network conditions, improving its robustness and availability. The cloud-edge collaborative architecture fully leverages the abundant computing resources of the cloud and the rapid response of the edge, achieving optimized allocation of computing resources. Model version management and automated deployment mechanisms ensure the security and traceability of system updates, providing a reliable guarantee for long-term stable operation.

[0024] It is not only suitable for monitoring hydropower station dams, but can also be extended to the safety monitoring of other large-scale infrastructure, such as bridge monitoring, tunnel monitoring, and high-rise building structure monitoring, and has broad application prospects and important engineering practical value. Attached Figure Description

[0025] Figure 1 The system architecture of this invention Figure One ; Figure 2 The system architecture of this invention Figure Two . Detailed Implementation

[0026] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] It should be noted that the directional concepts of "left", "right", "up", "down", "front", "back", "inner", and "outer" in the following scheme are all relative directions, and will not be listed one by one here.

[0028] Example 1: Construction and Training of Twin Brain Modules This embodiment details the specific structure and implementation of a twin brain module in a hydropower station dam safety monitoring data acquisition and transmission system. Deployed on a cloud server, this twin brain module serves as the intelligent core of the entire monitoring system. It possesses the ability to learn from multi-source historical monitoring data of the dam and establish a normal operation status representation model. It can also provide lightweight intelligent inference models for edge computing devices deployed on-site, thereby realizing a cloud-edge collaborative intelligent monitoring architecture. The design and implementation of this twin brain module provide strong technical support for the system described in this invention, enabling the entire monitoring system to possess highly intelligent data processing and analysis capabilities.

[0029] This twin brain module adopts a modular design, mainly comprising two core components: a sequence neural network engine and a reflection kernel generation module. These two components work together to achieve a complete process from learning from historical data to generating a lightweight model. The sequence neural network engine, as the core of intelligent learning, employs a neural network architecture with both parallel and recurrent equivalent representations. This architecture has significant technical advantages: during model training, it can fully utilize parallel computing resources, simultaneously inputting dam monitoring data from multiple time steps, including deformation data, seepage data, and stress-strain data reflecting the dam's structural response, as well as reservoir water level data, rainfall data, and ambient temperature data reflecting external environmental conditions. Parallel processing significantly improves training efficiency. During model inference, it automatically switches to a recurrent computing mode, recursively processing the input data step by step, achieving efficient real-time inference with constant-time complexity. This design makes the generated model particularly suitable for deployment and operation in resource-constrained edge computing environments.

[0030] The sequence neural network engine features a meticulously designed internal structure, including dedicated temporal mixing and channel mixing modules. These modules are responsible for handling temporal dependencies and multivariate coupling relationships in time series data, respectively. The temporal mixing module implements an advanced temporal information fusion mechanism. By maintaining a learnable temporal decay vector, it intelligently weights the multi-source historical monitoring data of the dam. Each element of this vector corresponds to a different time offset, enabling it to adaptively learn the importance distribution of historical data. This ensures that historical data further removed from the current time is assigned a smaller influence weight, effectively capturing long-term dependencies in dam monitoring data while avoiding the gradient vanishing problem common in traditional recurrent neural networks.

[0031] The time-mixing module employs an adaptive decay mechanism in its implementation. This mechanism not only considers the impact of time distance on data importance but also fully takes into account the inherent periodicity of dam monitoring data. The time decay weight is calculated using the following mathematical formula: ; in: Indicates the time offset as The decay weight; The basic attenuation coefficient controls the overall attenuation rate; The decay exponent determines the shape of the decay function; These are periodic modulation coefficients, controlling the intensity of periodic modulation; It is a periodic frequency parameter that determines the frequency of periodic changes; The time offset represents the time difference between historical data and the current moment. This formula introduces a periodic modulation term in the form of a sine function, enabling the time mixing module to effectively identify and utilize seasonal variation patterns, diurnal variation patterns, and other periodic physical phenomena in dam monitoring data. This is of great significance for accurately modeling the normal behavior patterns of dams.

[0032] The time fusion module employs a feature fusion mechanism when processing historical monitoring data. It weights and combines data from different time steps according to their time decay weights to generate feature representations containing rich temporal information. The mathematical expression for this feature fusion process is as follows: ; in: This is the feature representation vector after temporal mixing; Let be the input feature vector at the i-th time step; and The weight matrix is ​​learnable and adaptively adjusted through training; It is the bias vector; This represents element-wise multiplication; T is the total length of the time series. This design, by introducing the concepts of query and key, and drawing inspiration from attention mechanisms, enables the model to automatically learn the importance of data at different time steps, thereby achieving more intelligent time information fusion.

[0033] Corresponding to the time-mixing module, the channel-mixing module is specifically responsible for handling the complex coupling relationships between various physical quantities in the dam monitoring system. Because the safety state of a dam is influenced by a combination of factors, including the structural response characteristics and changes in external environmental conditions, complex nonlinear coupling relationships exist between these different monitoring variables. The channel-mixing module employs a multi-headed coupled attention mechanism to model these cross-variable correlations. This mechanism can automatically learn and capture potential correlation patterns between deformation data, seepage data, stress-strain data, and reservoir water level data, rainfall data, and ambient temperature data.

[0034] The channel mixing module employs a multi-head coupled attention mechanism to model cross-variable coupling relationships. The core idea of ​​this mechanism is to automatically discover and utilize the correlations between different monitored variables by learning the attention weights between them. Specifically, this mechanism calculates the attention weight of each monitored variable on all other variables, thereby achieving information interaction and fusion. The formula for calculating the attention weight is as follows: ; in: This represents the attention weight of variable i on variable j in the h-th attention head, reflecting the degree of influence of variable j on variable i; and The query vector and key vector are generated from the input features through a linear transformation, respectively. is a learnable coupling bias matrix used to encode prior physical coupling relationships; C is the total number of monitored variables; Let be the dimension of the key vector. The coupling bias matrix introduced in this formula is an important feature that allows the model to incorporate prior knowledge from domain experts to better learn the relationships between physically related variables.

[0035] The formula for calculating the output characteristics of the channel mixing module is as follows: Where: each attention head The calculation formula is: ; Intervariable interaction function Defined as: in: To output the weight matrix; H represents the output bias vector; H is the number of attention heads. Let h be the h-th value weight matrix; [;] denotes vector concatenation operation; For activation functions; This is the weight matrix of the interaction function. This interaction function can capture linear relationships, nonlinear interactions, and differential characteristics between variables.

[0036] The sequence neural network engine employs a gating mechanism to fuse temporal and channel features: ; ; in: This is the gate vector; This is the gate weight matrix; This is the gated bias vector; This is the final output feature.

[0037] The sequence neural network engine features a dual computation mode: during training, it employs a parallel mode to process historical monitoring data, improving training efficiency; during inference, it switches to a recurrent mode to achieve real-time prediction with constant-time complexity. The state update formula for the recurrent mode is: ;in: The current state; It is a dynamic forgetting factor.

[0038] Through deep learning of dam deformation data, seepage data, stress and strain data, reservoir water level data, rainfall data, and ambient temperature data, the sequence neural network engine generates a baseline twin model that can characterize the normal operating state of the dam.

[0039] The training of the baseline twin model employs a multi-task learning loss function: Among them: predicted loss Consistency loss Physical constraint loss is the weighting coefficient; N is the number of training samples; M is the number of consistency verification samples; K is the number of physical constraints; This is a predicted value; This is the actual value; For disturbance; This is the k-th physical constraint function; This is the constraint threshold.

[0040] The baseline twin model is trained using supervised learning, with monitoring data from historical normal operation as training samples. The model parameters are optimized by minimizing the error between the predicted value and the actual observed value.

[0041] The reflection kernel generation module is used to quantize and compress the trained benchmark twin model and optimize its parameters, extract the core inference logic, and generate a lightweight reflection kernel model with significantly reduced parameter size.

[0042] The reflection kernel generation module employs an adaptive importance pruning algorithm, and the parameter importance scoring formula is as follows: The formula for calculating the parameter retention probability is: in: The importance score for parameter p; p is a single parameter in the model; To validate the dataset Expectation operation on; The loss function; Let |p| be the gradient of the loss function with respect to parameter p; |p| is the absolute value of parameter p. Let p be the retention probability of parameter p; This is the pruning threshold coefficient; The median of the importance scores for all parameters; The importance score for parameter q; This is the set of all model parameters.

[0043] The lightweight objective function of the reflection kernel model is: in: This represents the overall lightweight loss of the reflection kernel model; For the sake of accuracy loss; This is due to dimensional loss; For delayed losses; This is the size loss weighting coefficient; For latency loss weighting coefficients; for accuracy loss ,in Output of the original model. For compressed model output, The square of the L2 norm; size loss ,in This refers to the dimensions of the compressed model. Target model size; delay loss ,in To reduce the inference latency of the compressed model, Delayed reasoning for the target.

[0044] The specific implementation steps of the reflection kernel generation module include: The baseline twin model is pruned and its parameters are quantized to remove redundant parameters and computational paths. Extract the core inference logic and key parameters of the benchmark twin model to generate a lightweight reflection kernel model suitable for edge computing environments; The generated reflection kernel model is subjected to edge device compatibility testing to ensure its stable operation in resource-constrained environments; The reflection kernel model is securely distributed to each edge neuron using encrypted transmission.

[0045] The reflection kernel model retains the time mixing and channel mixing mechanisms of the baseline twin model, but optimizes and simplifies the network structure, adopts a cyclic inference mode, and achieves real-time inference capability with constant time complexity.

[0046] Through the above technical solution, the twin brain module realizes the complete transformation from historical monitoring data of the dam to a lightweight reflective kernel model, providing intelligent support for the entire monitoring system and achieving an organic combination of cloud intelligence and edge computing.

[0047] Example 2: Deployment and Online Monitoring of Limbic Neurons Building upon Example 1, this example details the specific structure and implementation of an edge neuron in a hydropower station dam safety monitoring data acquisition and transmission system. This edge neuron, deployed as an intelligent terminal device at the dam monitoring site, undertakes key functions such as receiving lightweight inference models generated in the cloud, processing on-site monitoring data in real time, performing intelligent predictive analysis, calculating state deviation indicators, and implementing adaptive data transmission control. The edge neuron's design fully considers the complexity of the on-site environment and resource constraints, achieving efficient edge intelligent computing capabilities through a modular hardware and software architecture. The edge neuron includes core functional components such as a model and data interface module, a data preprocessing module, an online twin prediction module, a reflection deviation calculation module, a hierarchical transmission control module, and a local storage module. These modules work collaboratively to form a complete edge intelligent monitoring system, enabling highly intelligent data analysis and decision processing in resource-constrained edge environments, providing real-time, accurate, and reliable technical support for dam safety monitoring.

[0048] The model and data interface module, as the core interface component of the limbic neuron, undertakes a dual key function: on the one hand, it is responsible for communicating with the cloud-based twin brain module, receiving and managing lightweight reflection kernel models; on the other hand, it establishes data connections with various field sensor devices to achieve real-time acquisition and preprocessing of multi-source monitoring data. This module employs highly reliable communication protocols and data processing mechanisms to ensure the security of model transmission and the accuracy of data acquisition.

[0049] In terms of model reception and management, this module implements complete model lifecycle management functions. When the cloud-based twin brain module generates a new version of the reflection kernel model, the model and data interface module receives the model file through a secure network connection. To ensure the integrity and authenticity of the received model file, this module employs a rigorous integrity verification algorithm, the mathematical expression of which is: This verification algorithm combines cryptographic hash functions and cyclic redundancy check (CRC). It generates an integrity verification value by XORing the SHA256 hash of the received file with a timestamped CRC32 checksum. The verification process is based on the following criteria: ; in: The received model file data; To receive timestamps, used to prevent replay attacks; This indicates a bitwise XOR operation; The expected verification value pre-calculated and sent to the cloud; This is the difference between the receiving time and the sending time. This is a preset timeout threshold used to prevent the loading of expired models. A model file is considered valid and safe only when the validation values ​​match and the time difference is within the allowable range.

[0050] The rigorously validated reflection kernel model was securely loaded into the memory space of the limbic neurons and initialized. This module also manages model versions, maintaining information on currently active and historical versions, and supporting dynamic updates and rollback operations. Regarding data interfaces, this module establishes reliable connections with various types of field sensors, supporting multiple communication protocols and interface standards, including analog signal interfaces, digital communication interfaces, and industrial bus interfaces. Through these interfaces, the module can acquire real-time observed physical quantities of the dam structure and corresponding environmental variables. The observed physical quantities mainly include key parameters reflecting the dam structure's response characteristics, such as deformation, seepage flow, stress, and strain, which directly reflect the dam's current structural state and safety level. The observed environmental variables include external condition parameters affecting dam behavior, such as reservoir water level, ambient temperature, and rainfall, providing the intelligent prediction module with necessary environmental information.

[0051] The data preprocessing module is a key component of the limbic neuron, responsible for data cleaning, standardization, and format conversion. This module ensures that raw monitoring data from different sensors can be correctly understood and processed by the reflection kernel model. Due to the diversity of field sensor types and the different dimensions, numerical ranges, and sampling characteristics of the measured physical quantities, the data preprocessing module transforms these heterogeneous raw data into standardized input in a unified format through a series of carefully designed processing steps. This module includes three core processing units: an instance normalization unit, a sequence segmentation unit, and a token generation unit. These three units work collaboratively in a pipeline manner to achieve a complete transformation process from raw observation data to model input tokens.

[0052] The instance normalization unit, as the first step in data preprocessing, is specifically responsible for addressing the issue of inconsistent data scales between different sensors. This unit independently normalizes the observation sequences for each sensor channel, effectively eliminating dimensional and numerical range differences between sensors and ensuring that data from different physical quantities can be processed at the same numerical scale. The instance normalization unit employs a dynamic window adaptive strategy, automatically adjusting the normalization parameters based on real-time data changes. Its core algorithm formula is: ; ; The normalized output is: ; in: Let be the dynamic mean of the i-th sensor at time t; For the i-th sensor at the th The dynamic mean at any given time; The coefficients are dynamically updated to control the degree to which historical information is retained; W is the length of the sliding window. For from the first The summation operation from time t to time t; Let be the observation value of the i-th sensor at time k; Let be the dynamic standard deviation of the i-th sensor at time t; For the i-th sensor at the th The dynamic standard deviation at time point; Let be the normalized output value of the i-th sensor at time t; Let be the original observation value of the i-th sensor at time t; To prevent division by zero of small constants; Let be the learnable scaling parameter for the i-th sensor; Let be the learnable translation parameters of the i-th sensor.

[0053] The sequence segmentation unit divides the long sequence into short sequence blocks suitable for processing by the reflection kernel model, according to a fixed block length and overlap step. The token generation unit converts each sequence block into a standardized input token containing positional and feature encodings.

[0054] The online twin prediction module calls the cyclic inference mode of the reflection kernel model. Based on the environmental variable observation values ​​and historical state information at the current moment, it predicts the theoretical normal values ​​of each monitored physical quantity under the current working condition in real time through a cyclic inference method with constant time complexity, and constructs the sequence of theoretical normal values ​​into a dynamic reference trajectory.

[0055] The prediction algorithm of the online twin prediction module adopts an environment-adaptive state update mechanism: The formula for calculating the predicted output is: in: For predicting the state; This is the dynamic forgetting coefficient; A vector of environment variables; It is an environment-state mapping function; This is the theoretical normal value; This is the weight matrix; This is the bias vector.

[0056] The reflection deviation calculation module calculates the Euclidean distance between the actual observed value and the theoretical normal value of the physical quantity as the reflection deviation value, and includes a deviation statistical analysis unit for calculating the statistical characteristics of the reflection deviation value within a sliding time window, including mean, variance, trend coefficient, and cumulative characteristics.

[0057] The reflection deviation is calculated using the Euclidean distance algorithm. ;in: This is the reflection deviation value; This is a vector of actual observed values ​​of physical quantities; This is the theoretical normal value vector.

[0058] The deviation statistical analysis unit calculates statistical characteristics within a preset sliding time window: Average deviation: Bias Variance: Where: W is the length of the sliding window; The average deviation; This represents the deviation and variance.

[0059] The hierarchical transmission control module executes a three-level adaptive data transmission strategy based on the magnitude, duration, and rate of change of the reflection deviation value: Level 1 local recording response: Triggered when the reflection deviation value is lower than the first preset threshold, the reflection deviation value and the corresponding observation value are stored in the local storage module; Second-level summary information transmission response: Triggered when the reflection deviation value is continuously higher than the first preset threshold but lower than the second preset threshold, a data summary containing deviation statistical characteristics, change trend analysis and time stamp is generated and transmitted. Level 3 Real-time Early Warning Information Transmission Response: Triggered when the reflection deviation value instantaneously exceeds the second preset threshold or the rate of change exceeds the preset limit, immediately intercepting the complete data record of the abnormal period for real-time transmission.

[0060] The local storage module is used to cache observation data, reflection deviation values, and system operation logs. This module employs a multi-tiered storage architecture, including a cache, a local database, and backup storage, to ensure data reliability and traceability.

[0061] The hardware architecture of the limbic neuron includes: Embedded computing unit: Employs a low-power, high-performance processor and pre-set memory capacity to run reflection kernel models and data processing tasks; Multi-channel data acquisition unit: supports simultaneous input of analog signals, digital signals and pulse signals; Local storage unit: Provides a preset capacity of data cache space, supporting local data backup and historical query; Communication unit: Supports wireless network communication and wired network communication, and has communication redundancy and fault switching capabilities; Clock synchronization unit: Achieves high-precision time synchronization through satellite positioning system to ensure time consistency of distributed monitoring data.

[0062] Through the above technical solution, the edge neuron realizes intelligent data analysis and decision-making in resource-constrained edge environments, can calculate reflection deviation values ​​in real time and execute hierarchical transmission strategies, and provides reliable protection for dam safety monitoring.

[0063] Example 3: Implementation of the Closed-Loop Model Evolution Mechanism This embodiment details the construction and implementation of a closed-loop model evolution mechanism in a hydropower station dam safety monitoring data acquisition and transmission system. As the intelligent core component of the entire monitoring system, this closed-loop model evolution mechanism possesses the ability to learn autonomously and continuously optimize. It can continuously learn new knowledge and patterns from the actual data generated during system operation and feed this knowledge back into the system, achieving continuous improvement in the monitoring system's performance. The core concept of this mechanism is to establish a complete closed-loop feedback system from edge perception to cloud intelligence and then to edge deployment, achieving system self-evolution and performance optimization through a data-driven approach.

[0064] The closed-loop model evolution mechanism is designed to fully consider the long-term and complex nature of dam safety monitoring. During long-term monitoring, the dam's behavior patterns may slowly change, new anomaly patterns may emerge, and external environmental conditions may change, all of which require the monitoring system to possess adaptability and learning capabilities. This mechanism establishes a complete data feedback channel to collect various types of data generated by edge neurons during actual operation, including normal operation data, anomaly detection data, and system response data. Then, advanced machine learning techniques are used to deeply analyze and mine this data, extracting valuable information for model optimization.

[0065] The closed-loop model evolution mechanism comprises four main components: a data collection and feedback subsystem, an incremental learning processing subsystem, a model version management subsystem, and an automated deployment subsystem. The data collection and feedback subsystem receives operational data from each edge neuron. When the hierarchical transmission control module of the edge neuron detects a need to upload data, it transmits the corresponding data to the cloud-based twin brain module via the network. This data mainly includes two categories: one is the data summary generated by the second-level summary information transmission response, containing key information such as deviation statistical characteristics, trend analysis, and time stamps; the other is the complete data record generated by the third-level real-time early warning information transmission response, containing detailed monitoring data and system response information for abnormal periods.

[0066] The incremental learning processing subsystem is the core component of the entire closed-loop mechanism. This subsystem receives the feedback data and processes it intelligently. First, it performs a comprehensive quality assessment of the newly collected data to ensure that the data used for learning has sufficient quality and reliability. Then, the system performs intelligent labeling, classifying the collected data according to the event types they represent, primarily into three categories: normal changes, abnormal events, and equipment failures. Next, the system uses advanced incremental training algorithms to update the parameters and optimize the performance of the existing baseline Siamese model, learning new behavioral patterns and abnormal features while maintaining historical knowledge retention.

[0067] The incremental learning module, as the intelligent processing core of the closed-loop model evolution mechanism, plays a crucial role in extracting valuable information from the raw backflow data and using it for model optimization. This module first conducts a comprehensive quality assessment of the data backflowed from the edge neurons, establishing a complete multi-dimensional data quality inspection system. The data quality assessment process includes four main dimensions: integrity check ensures that the uploaded data contains all necessary sensor readings and environmental variable information, avoiding the impact of missing data on learning performance; consistency verification identifies potential sensor malfunctions or abnormal readings by cross-validating the data correlations between different sensors; temporal continuity analysis checks for missing, duplicate, or abnormal jumps in the data's timestamp sequence, ensuring the temporal logic correctness of the data; and noise level assessment quantifies the random disturbance components in the data through statistical analysis methods, filtering out low-quality data with excessive noise.

[0068] The overall data quality score employs a multi-dimensional weighted evaluation algorithm, the mathematical expression of which is: ; in: This is the overall quality score, with a value ranging from 0 to 1; For dataset The score on the i-th quality dimension; The weight coefficient for the i-th dimension reflects the importance of that dimension to the overall quality. To adjust parameters and control the degree of penalty for deviations from the baseline value; Let be the baseline value for the i-th dimension, representing the expected quality level. This formula ensures through a product form that serious quality problems in any dimension will significantly affect the overall score, while an exponential term applies additional penalties for deviations from the baseline value.

[0069] After the data quality assessment is passed, the incremental learning module performs intelligent labeling, a crucial step in data preprocessing. The labeling process employs a semi-supervised learning method, combining domain expert knowledge and automated analysis algorithms to accurately classify the collected data according to the event types they represent. The system categorizes events into three main types: normal changes, including changes in dam behavior caused by environmental conditions (such as seasonal temperature variations, water level fluctuations, and climate change), which conform to physical laws and are within predictable limits; abnormal events, including events that may affect dam safety but do not yet pose a direct threat, such as slight accelerated structural deformation, abnormally increased seepage flow, and changes in stress distribution; and equipment failures, including sensor malfunctions or problems with the measurement system, such as sensor drift, communication interruptions, and abnormal data acquisition.

[0070] The incremental learning module employs an advanced incremental training algorithm to intelligently update the parameters of the baseline Siamese model. The core objective of this algorithm is to maintain the retention of historical knowledge while learning new knowledge, effectively avoiding catastrophic forgetting. The incremental training process uses an elastic weight consolidation strategy, calculating the importance of model parameters to historical task performance and setting different update constraints for different parameters. The mathematical expression of its incremental training loss function is: ; in: The loss for supervised learning on new data; The importance weight of parameter i controls the update magnitude of this parameter; Let i be the i-th diagonal element of the Fisher information matrix, representing the sensitivity of the quantization parameter i to the model output. This refers to the i-th parameter of the current model; These are the key parameter values ​​obtained from training on historical data. The loss function uses a regularization term to ensure that key parameters do not deviate too far from their historical optimum, thus maintaining the model's performance on historical tasks.

[0071] The incremental learning module also employs an experience replay strategy to further enhance learning performance. This strategy intelligently mixes newly collected data with samples selected from a historically representative database during training, effectively preventing overfitting of the model to new data. The mixed training batches are constructed using the following formula: in: For the final batch of mixed data used for training; Training batches consisting of newly collected data; To maintain a historically representative database; This represents the number of replay samples taken from the historical database. The experience-based replay strategy uses an intelligent sampling algorithm to select the most representative and important historical samples for replay, ensuring that the model does not forget important historical experiences while learning new knowledge.

[0072] To further enhance the model's ability to identify abnormal patterns, the incremental learning module employs a contrastive learning approach, constructing positive and negative sample pairs to improve the model's discriminative performance. Contrastive learning enables the model to better distinguish between normal and abnormal states, improving the accuracy and robustness of anomaly detection. This module also establishes a multi-level model performance evaluation system, including three main levels: prediction accuracy, anomaly detection rate, and false positive rate. Each level contains multiple specific evaluation metrics to ensure the quality and reliability of model updates.

[0073] The model version management module is responsible for maintaining a complete model iteration record. This module adopts a distributed version control design, generating a unique identifier for each model version and recording detailed change information, including model structure parameters, performance metrics, training data summaries, and change descriptions. When the newly trained model's comprehensive score in the multi-level performance evaluation system reaches the preset improvement standard, the system automatically triggers the version release process. The comprehensive performance evaluation uses a hierarchical weighted scoring mechanism, and its calculation formula is as follows: ;in: The overall performance score; These are the weight coefficients for the l-th layer; This represents the score for layer l. The three layers correspond to the prediction accuracy index, anomaly detection rate index, and false alarm rate index, respectively, and the weight of each layer is adjusted according to the actual application requirements.

[0074] Once the model version management module confirms the release of a new version, the automated deployment subsystem initiates the generation and distribution process of the reflection kernel model. This process first calls the reflection kernel generation module to quantize, compress, and optimize the parameters of the updated baseline twin model, generating a lightweight reflection kernel model suitable for edge computing environments. The model distribution process employs an incremental update mechanism. By calculating the parameter differences between the new version and the current deployment version, only the changed parameters are transmitted, significantly reducing the amount of data transmitted over the network and improving update efficiency. The distribution process uses encrypted transmission and digital signature technology to ensure the security and integrity of the model files during transmission.

[0075] To ensure the effectiveness and security of the system's continuous self-evolution capability, the closed-loop model evolution mechanism also implements multiple safeguards. The data quality monitoring mechanism continuously monitors the quality trends of uploaded data, identifies and filters low-quality data, and prevents poor-quality data from negatively impacting model training. The model performance monitoring mechanism continuously tracks the performance of deployed models in actual operation, analyzes the trends of performance indicators to promptly detect performance degradation issues, and automatically triggers model rollback and retraining processes when abnormal performance declines are detected. The security assurance mechanism ensures that model updates do not introduce security risks or privacy leaks through multiple verification and audit processes, including model behavior consistency checks, security vulnerability scanning, and privacy protection assessments.

[0076] Through the above complete technical solution, the closed-loop model evolution mechanism successfully realizes the continuous self-evolution capability of the hydropower station dam safety monitoring system, enabling the entire system to continuously learn new knowledge and patterns during long-term operation, continuously improve monitoring accuracy, early warning capability and system reliability, and provide strong technical support for the long-term safe operation of the dam.

[0077] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A data acquisition and transmission system for safety monitoring of hydropower station dams, characterized in that, include: The twin brain module, deployed on a cloud server, includes: The sequence neural network engine adopts a neural network architecture with dual equivalent representations of parallelism and recurrence. The architecture includes a time mixing module and a channel mixing module. The time mixing module performs weighted processing on the multi-source historical monitoring data of the dam through a learnable time decay vector. The channel mixing module is used to realize cross-variable coupling relationship modeling. During training, the sequence neural network engine uses a parallel mode to process the deformation data, seepage data, stress and strain data of the dam, as well as reservoir water level data, rainfall data, and ambient temperature data, to generate a benchmark twin model that can characterize the normal operating state of the dam. The reflection kernel generation module is used to quantize and compress the trained benchmark twin model and optimize its parameters, extract the core inference logic, generate a lightweight reflection kernel model with significantly reduced parameter size, and distribute it to the edge neurons via network communication. And at least one edge neuron deployed at the dam monitoring site, including: The model and data interface module is used to receive and load the reflection kernel model from the cloud, establish a data connection with the field sensors, and obtain the actual observed values ​​of physical quantities measured by the sensors and the corresponding environmental variable observation values ​​in real time. The data preprocessing module includes an instance normalization unit, a sequence segmentation unit, and a token generation unit. It is used to standardize the continuous observation sequence, segment it into blocks according to a preset length and step size, and convert the segmented data into block sequence tokens that conform to the input format of the reflection kernel model. The online twin prediction module calls the cyclic inference mode of the reflection kernel model. Based on the environmental variable observation values ​​and historical state information at the current moment, it predicts the theoretical normal values ​​of each monitored physical quantity under the current working condition in real time through a cyclic inference method with constant time complexity, and constructs the sequence of theoretical normal values ​​into a dynamic reference trajectory. The reflection deviation calculation module calculates the Euclidean distance between the actual observed value and the theoretical normal value of the physical quantity as the reflection deviation value, and includes a deviation statistical analysis unit for calculating the statistical characteristics of the reflection deviation value within a sliding time window, including mean, variance, trend coefficient and cumulative characteristics. The hierarchical transmission control module executes a three-level adaptive data transmission strategy based on the magnitude, duration, and rate of change of the reflection deviation value: when the reflection deviation value is lower than the first preset threshold, a first-level local recording response is triggered; when the reflection deviation value is continuously higher than the first preset threshold but lower than the second preset threshold, a second-level summary information transmission response is triggered; and when the reflection deviation value is instantaneously higher than the second preset threshold or the rate of change exceeds a preset limit, a third-level immediate warning information transmission response is triggered. The local storage module is used to cache observation data, reflection deviation values, and system operation logs; The system also includes a closed-loop model evolution mechanism. The incremental learning module receives the uploaded summary information and early warning information data, performs incremental training and parameter fine-tuning on the benchmark twin model, and maintains the iteration record through the model version management module. After the model is updated, the generation and distribution of the new version of the reflection kernel model are automatically triggered, realizing the system's continuous self-evolution capability.

2. The system according to claim 1, characterized in that, The temporal mixing module of the sequence neural network engine assigns different decay weights to data at different time steps in the historical monitoring data sequence, so that the influence weight of historical data further away from the current time is smaller, thereby effectively capturing long-term dependencies. The channel mixing module achieves correlation modeling between different monitoring variables by interacting and fusing information on the feature dimension; the sequence neural network engine adopts a parallel computing mode to improve training efficiency during the training phase and switches to a loop computing mode during the inference phase to achieve efficient real-time inference.

3. The system according to claim 1, characterized in that, The workflow of the reflection kernel generation module includes: pruning and parameter quantization of the trained baseline twin model to remove redundant parameters and computational paths; extracting the core inference logic and key parameters of the baseline twin model to generate a lightweight reflection kernel model suitable for edge computing environments; performing edge device compatibility testing on the generated reflection kernel model to ensure its stable operation in resource-constrained environments; and securely distributing the reflection kernel model to each edge neuron via encrypted transmission.

4. The system according to claim 1, characterized in that, The prediction mechanism of the online twin prediction module includes: receiving the current reservoir water level observation value, ambient temperature observation value, and rainfall observation value as input conditions; calculating the expected value of each monitored physical quantity at the next moment through cyclical inference of the reflection kernel model based on the environmental variable observation value and historical state information; forming a dynamic reference trajectory from the continuously predicted expected value sequence, which reflects the normal behavior pattern of the dam under the current environmental conditions; the prediction process has constant time complexity to ensure real-time response requirements.

5. The system according to claim 1, characterized in that, The reflection deviation calculation module quantifies the degree of system deviation by calculating the Euclidean distance between the actual observed value vector and the theoretical normal value vector; the deviation statistical analysis unit continuously tracks the change of reflection deviation value within a preset sliding time window and calculates its statistical characteristics, including average deviation, deviation variance, trend coefficient, and cumulative deviation, to assess the stability and abnormal development trend of the system state.

6. The system according to claim 1, characterized in that, The three-level response strategy of the hierarchical transmission control module is as follows: The first level is local recording response. When the reflection deviation value is lower than the first preset threshold, the reflection deviation value and the corresponding observation value are stored in the local storage module without data transmission to save communication resources; the second level is summary transmission response. When the reflection deviation value continuously exceeds the first preset threshold but is lower than the second preset threshold, a data summary containing deviation statistical characteristics, trend analysis and time stamp is generated and transmitted over the network in a low-priority manner. The third-level early warning transmission response immediately intercepts the complete data record of the abnormal period when the reflection deviation value exceeds the second preset threshold or the deviation change rate exceeds the preset safety limit, and transmits it in real time with the highest priority.

7. The system according to claim 1, characterized in that, The instance normalization unit of the data preprocessing module performs independent normalization processing on the observation sequence of each sensor channel to eliminate the differences in dimensions and numerical ranges between different sensors. The sequence segmentation unit divides the long sequence into short sequence blocks suitable for reflection kernel model processing by dividing the long sequence into blocks with fixed block length and overlap step size; the token generation unit converts each sequence block into a standardized input token containing position encoding and feature encoding.

8. The system according to claim 1, characterized in that, The closed-loop model evolution mechanism achieves continuous learning through the following steps: the hierarchical transmission control module uploads the data generated by the second and third level responses to the twin brain module; the incremental learning module performs quality assessment and labeling on the collected new data, distinguishing between normal changes, abnormal events, and equipment failures; and the incremental training algorithm is used to update the parameters of the baseline twin model, maintaining the memory of historical knowledge while learning new behavioral patterns. The model version management module records the performance metrics and changes for each update, and triggers a version release when the performance of the new model reaches the preset improvement standard.

9. The system according to claim 8, characterized in that, The incremental learning module employs an experience replay strategy, mixing newly collected data with historical representative data in a preset ratio for training to prevent the model from overfitting to new data; it uses a contrastive learning method to enhance the model's ability to identify abnormal patterns, improving the model's discriminative performance by constructing positive and negative sample pairs; and it sets up a multi-level model performance evaluation system, including prediction accuracy indicators, anomaly detection rate indicators, and false alarm rate indicators, to ensure the quality and reliability of model updates.

10. The system according to claim 1, characterized in that, The hardware architecture of the edge neuron includes: an embedded computing unit, employing a low-power, high-performance processor and a preset memory capacity, for running the reflection kernel model and data processing tasks; a multi-channel data acquisition unit, supporting simultaneous access of analog, digital, and pulse signals; a local storage unit, providing a preset data cache space to support local data backup and historical retrieval; a communication unit, supporting wireless network communication and wired network communication, with communication redundancy and fault switching capabilities; and a clock synchronization unit, achieving high-precision time synchronization through a satellite positioning system to ensure the time consistency of distributed monitoring data.

Images