A water conservancy project safety monitoring method and system based on data processing

Abstract

The application relates to the technical field of data processing, in particular to a water conservancy project safety monitoring method and system based on data processing. The method comprises the following steps: constructing a multi-source data acquisition scheme for a water conservancy project to obtain a multi-source data acquisition scheme; collecting multi-source data of the water conservancy project according to the multi-source data acquisition scheme to obtain an original data stream; classifying monitoring nodes according to the multi-source data acquisition scheme to obtain an initial node set; constructing and defining an initial edge according to the initial node set to obtain an initial edge list; assigning weights to different types of initial edges of the initial edge list and generating an initial graph structure to obtain the initial graph structure; and performing correlation analysis on the original data stream and adjusting the weights of dynamic edges of the initial graph structure to obtain a dynamic topology graph. The application realizes the intellectualization, automation and refinement of water conservancy project safety monitoring through data processing technology, and significantly improves the engineering safety guarantee level.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a water conservancy project safety monitoring method and system based on data processing. BACKGROUND

[0002] The water conservancy project itself has a complex structure, a variable operating environment, and is long-term affected by factors such as water flow, sediment, and temperature changes, which can easily cause various safety hazards such as structural deformation, seepage, and landslides. Therefore, safety monitoring of water conservancy projects to timely detect and warn potential risks is of great significance to the safety of the project and the safety of people's lives and property.

[0003] Traditional methods usually use simple threshold alarms, that is, when a certain monitoring parameter exceeds the preset threshold, an alarm is issued, so it is difficult to handle complex data correlations and has insufficient warning capability.

[0004] The disadvantages of being difficult to handle complex data correlations are as follows: the water conservancy project is a complex system, and there are complex interactions and influences between various monitoring parameters. For example, increased rainfall causes the reservoir water level to rise, which in turn affects the dam seepage pressure. Isolated analysis of a single parameter can easily cause misjudgment and missed judgment, ignoring the interaction between parameters. The safety state of the water conservancy project is related not only to the current monitoring data, but also to the historical data and spatial position. For example, a sudden displacement of a monitoring point indicates a risk, but if the displacement data of surrounding monitoring points are analyzed, it will be found that this is only a local disturbance and does not constitute a risk. The traditional method is difficult to capture this spatio-temporal correlation.

[0005] The disadvantages of insufficient warning capability are as follows: the setting of the threshold needs to consider various factors such as project type, operating conditions, and environmental conditions. A fixed threshold is difficult to adapt to complex and variable actual situations, and is prone to false alarms and missed alarms. The traditional method can only analyze the current monitoring data and cannot predict future risk trends. This makes the warning lag and makes it difficult to take preventive measures in advance. SUMMARY

[0006] Therefore, it is necessary to provide a water conservancy project safety monitoring method and system based on data processing to solve at least one of the above technical problems.

[0007] To achieve the above-mentioned purpose, a water conservancy project safety monitoring method based on data processing comprises the following steps:

[0008] Step S1: constructing a multi-source data acquisition scheme for the water conservancy project to obtain a multi-source data acquisition scheme; collecting multi-source data of the water conservancy project according to the multi-source data acquisition scheme to obtain an original data stream;

[0009] Step S2: hierarchical monitoring nodes are obtained according to a multi-source data acquisition scheme; an initial edge list is obtained according to the initial node set; different types of initial edge weight assignment are performed on the initial edge list, and an initial graph structure is generated to obtain the initial graph structure; correlation analysis is performed on the original data stream to obtain a monitoring point correlation matrix; and the initial graph structure is adjusted by using the monitoring point correlation matrix to obtain a dynamic topology graph;

[0010] Step S3: obtain water conservancy project prediction demand data; construct an ST-GCN model architecture according to the water conservancy project prediction demand data to obtain a model structure configuration; construct an ST-GCN model according to the model structure configuration and the dynamic topology graph, and perform ST-GCN model training to obtain an ST-GCN model;

[0011] Step S4: use the ST-GCN model to predict the future state to obtain a prediction state sequence; perform risk preliminary judgment on the prediction state sequence to obtain individual risk assessment results; perform multi-parameter comprehensive risk assessment based on a Bayesian network according to the individual risk assessment results to obtain comprehensive risk assessment results; generate a risk assessment report according to the comprehensive risk assessment results to obtain a risk assessment report; judge the risk assessment report according to the pre-warning trigger condition, and publish the pre-warning information to realize the water conservancy project safety monitoring task.

[0012] The application realizes comprehensive collection, format unification and quality improvement of multi-source heterogeneous data of water conservancy projects by constructing a multi-source data collection scheme and collecting and preprocessing data according to the scheme, and lays a solid data foundation for subsequent analysis. The problems of single data source and uneven data quality in traditional methods are effectively solved, and the reliability and usability of monitoring data are improved. By constructing a dynamic topology graph, the structural information of water conservancy projects, the spatial relationship between monitoring points and the time sequence correlation of monitoring data are organically combined to provide more expressive and informative input data for subsequent deep learning models. The problem of complex data correlation that is difficult to handle by traditional methods is effectively solved, so that the model can better learn the operation rules and risk patterns of water conservancy projects. By constructing and optimizing the ST-GCN model, the model can effectively learn the spatio-temporal features in the dynamic topology graph and accurately predict the future state of the water conservancy project. The spatio-temporal convolution structure, attention mechanism and parameter optimization strategy of the model ensure the prediction accuracy and generalization ability of the model, and provide reliable prediction results for risk assessment. The trained ST-GCN model is used to predict the future state, and the Bayesian network is used for multi-parameter comprehensive risk assessment, realizing comprehensive and accurate assessment of the risk of water conservancy projects, and automatically warning according to the preset warning strategy, finally forming a risk assessment report and publishing warning information. This step effectively solves the problems of strong subjectivity and lagging warning of traditional risk assessment methods, and realizes the intelligentization and automation of water conservancy project safety monitoring. Therefore, the application provides a water conservancy project safety monitoring method based on data processing, effectively solves the drawbacks of traditional water conservancy project safety monitoring methods, realizes the intelligentization, automation and refinement of safety monitoring, and significantly improves the engineering safety guarantee level.

[0013] Preferably, step S1 comprises the following steps:

[0014] Step S11: sensor deployment is performed on the water conservancy project, and a data transmission network is constructed to obtain a multi-source data collection scheme;

[0015] Step S12: multi-source data of the water conservancy project is collected according to the multi-source data collection scheme to obtain real-time multi-source data flow;

[0016] Step S13: data format conversion and verification are performed on the real-time multi-source data flow to obtain formatted and verified data;

[0017] Step S14: multi-source data integration and time synchronization are performed on the formatted and verified data to obtain an original data stream.

[0018] The multi-source data acquisition scheme obtained through detailed sensor layout planning for water conservancy projects and the construction of a reliable data transmission network can comprehensively cover key monitoring areas, ensuring the completeness and representativeness of the collected data, and providing a reliable data foundation for subsequent analysis and evaluation. The application of LoRaWAN technology reduces the power consumption and cost of data transmission, realizes remote monitoring and real-time data collection, and improves monitoring efficiency. Clear sensor models and installation specifications ensure data collection accuracy and consistency, avoiding data errors caused by equipment differences or improper installation. According to the pre-prepared multi-source data acquisition scheme, data from different types of sensors can be obtained, forming real-time multi-source data streams that reflect the real-time operation status of water conservancy projects. The pre-set sampling frequency can capture dynamic changes in project status, and according to the characteristics of different parameters, sampling resources are reasonably allocated to avoid data redundancy or information loss. Real-time data transmission and storage provide a guarantee for timely detection of abnormal conditions and rapid response. Through data format conversion and verification of real-time multi-source data streams, raw data collected by different sensors can be converted into unified physical units and standard formats, facilitating subsequent data processing and analysis. Strict data verification rules, such as range checks and outlier detection, can effectively identify and eliminate incorrect data, improving data quality and ensuring the accuracy and reliability of analysis results. The automated data verification process reduces the workload of manual intervention and improves data processing efficiency. Through multi-source data integration and time synchronization of formatted and verified data, data from different sensors and different sampling frequencies can be aligned to a unified time axis to form a complete and time-synchronized raw data stream. The application of linear interpolation methods effectively solves the problem of data misalignment caused by different sampling frequencies, ensuring the accuracy of data analysis. Data integration and time synchronization provide necessary data preparation for subsequent construction of dynamic topology atlas and spatiotemporal analysis, laying a foundation for training and application of deep learning models.

[0019] Preferably, step S2 comprises the following steps:

[0020] Step S21: Obtain a water conservancy project structure design drawing; map the multi-source data acquisition scheme and the water conservancy project structure design drawing to obtain a node mapping set;

[0021] Step S22: Classify the node importance of the node mapping set to obtain an initial node set;

[0022] Step S23: Construct and define the initial edges according to the water conservancy project structure design drawing and the initial node set to obtain an initial edge list; assign weights to different types of initial edges in the initial edge list to obtain a weighted initial edge list;

[0023] Step S24: generating an initial graph structure according to the weighted initial edge list and the initial edge list;

[0024] Step S25: performing correlation analysis on the original data stream based on historical data to obtain a monitoring point correlation matrix; and performing original data preprocessing on the original data stream to obtain a preprocessed data stream;

[0025] Step S26: adjusting the dynamic edge weight of the initial graph structure by using the monitoring point correlation matrix to obtain a dynamic weighted graph;

[0026] Step S27: performing data fusion on the dynamic weighted graph and the preprocessed data stream to obtain a dynamic topology graph.

[0027] The application maps sensor information in a multi-source data acquisition scheme to a water conservancy engineering structure design drawing, establishes a spatial correspondence between sensor data and engineering structures, links abstract sensor data to specific engineering parts, and provides a basis for subsequent graph structure construction and spatial analysis. Accurate monitoring point three-dimensional coordinate information can accurately reflect the position of the monitoring point in the engineering structure, improving the accuracy of subsequent analysis. The establishment of the node mapping set provides necessary data support for subsequent node importance classification and edge connection relationship determination. Importance classification of nodes highlights the role and position of different monitoring points in engineering safety monitoring, enabling the model to focus more on monitoring data from critical parts, improving the prediction accuracy of the model and the accuracy of risk assessment. The graded initial node set provides a basis for subsequent construction of more targeted graph structures and differentiated data analysis. By combining the water conservancy engineering structure design drawing and the initial node set, an initial edge list containing multiple edge types is constructed, which can more comprehensively reflect the spatial relationship and mutual influence between monitoring points. Different types of edge weight assignment, such as weight assignment based on material strength and distance, make the edge weight more physically meaningful and more accurately express the correlation strength between monitoring points, providing more effective information for subsequent graph convolution operations and feature learning. Using the weighted initial edge list and the initial node set, an initial graph structure of the water conservancy engineering is constructed, which abstracts the engineering structure into a graph model, providing a data structure foundation for subsequent graph convolution operations and deep learning. This graph structure not only contains the spatial position information of the monitoring points, but also contains the connection relationship and correlation strength between the monitoring points, providing an important information carrier for the model to learn the structural characteristics and state change rules of the water conservancy engineering. Through correlation analysis of the original data stream based on historical data, a monitoring point correlation matrix is obtained, which quantifies the data correlation between different monitoring points, providing a basis for subsequent dynamic adjustment of edge weights. At the same time, the original data stream is preprocessed, including outlier removal, missing value filling and data normalization, improving the quality and consistency of the data, and providing more reliable input data for subsequent model training and prediction. Using the monitoring point correlation matrix to dynamically adjust the edge weights of the initial graph structure enables the graph structure to reflect the real-time data correlation between monitoring points, improving the sensitivity and adaptability of the model to engineering state changes. The dynamically adjusted edge weights enable the graph structure to better capture abnormal changes in the engineering state, providing more accurate information for risk assessment and early warning. The dynamic weighted graph and the preprocessed data stream are fused to generate a dynamic topology atlas containing spatiotemporal information, providing high-quality input data for the training and prediction of the ST-GCN model. The dynamic topology atlas not only contains the topological information of the engineering structure, but also contains the real-time monitoring data of each monitoring point, enabling the model to simultaneously learn the static characteristics and dynamic change rules of the engineering structure, thereby more accurately predicting future states and assessing risks.

[0028] Preferably, step S23 comprises the following steps:

[0029] Step S231: marking physical connection edges for the initial node set to obtain physical connection edges; marking spatial proximity edges for the initial node set to obtain spatial proximity edges;

[0030] Step S232: marking water flow direction associated edges for the initial node set to obtain water flow associated edges; marking force transmission path associated edges for the initial node set to obtain force associated edges;

[0031] Step S233: constructing an initial edge list according to the physical connection edges, the spatial proximity edges, the water flow associated edges, and the force associated edges to obtain the initial edge list;

[0032] Step S234: assigning weights to the physical connection edges based on a material strength parameter to obtain physical connection edge weights; assigning weights to the spatial proximity edges based on an inverse of a Euclidean distance between nodes to obtain spatial proximity edge weights;

[0033] Step S235: assigning weights to the water flow associated edges based on a fixed weight value of a directed edge to obtain water flow direction associated edge weights; assigning weights to the force associated edges based on a stress transmission coefficient to obtain force transmission path associated edge weights;

[0034] Step S236: constructing a weighted initial edge list according to the physical connection edge weights, the spatial proximity edge weights, the water flow direction associated edge weights, and the force transmission path associated edge weights to obtain the weighted initial edge list.

[0035] This invention clarifies two basic connection relationships between monitoring points by marking physical connection edges and spatial proximity edges. Physical connection edges reflect the inherent connection characteristics of the engineering structure, while spatial proximity edges consider the influence of spatial distance between monitoring points. The combined effect of these two types of edges provides a more complete description of the spatial correlation between monitoring points, laying the foundation for subsequent graph structure construction and spatial analysis. By marking water flow direction-related edges and force transmission path-related edges, the influence of water flow and stress transmission paths is introduced, enabling the graph structure to more accurately reflect the physical characteristics and mechanical behavior of hydraulic engineering projects. Water flow-related edges reflect the influence of seepage, while force transmission-related edges reflect the influence of stress transmission; this information is crucial for understanding changes in engineering status and risk assessment. Integrating different types of edges into an initial edge list constructs a complete list containing various connection relationships, providing a foundation for subsequent weight assignment and graph structure generation. This list clearly records the various connection relationships between monitoring points, providing important topological information for subsequent analysis and modeling. Weighting physical connection edges based on material strength parameters allows the edge weights to reflect the material properties of the connection points; for example, materials with higher strength correspond to larger edge weights, more accurately expressing the strength of the physical connection. Weighting spatially adjacent edges based on the reciprocal of Euclidean distance considers the impact of spatial distance on the correlation between monitoring points; the closer the distance, the greater the weight, consistent with the general rules of spatial correlation. Assigning fixed weights to flow-related edges simplifies the modeling process of flow influence while ensuring the representation of flow direction information in the graph structure. Weighting stress-related edges based on stress transfer coefficients allows edge weights to reflect the efficiency of stress transfer between monitoring points, thus more accurately expressing the mechanical correlation between them. Integrating the weight information of different edge types, a weighted initial edge list is constructed, providing complete edge information, including connectivity and weights, for the subsequent generation of the initial graph structure. This list comprehensively considers multiple factors such as physical connectivity, spatial proximity, flow influence, and stress transfer paths, enabling the initial graph structure to more comprehensively and accurately reflect the actual situation of the hydraulic engineering project.

[0036] Preferably, step S26 includes the following steps:

[0037] Step S261: Initialize the weights of the initial graph structure based on statistical correlation using the monitoring point correlation matrix to obtain a preliminary dynamic weight graph;

[0038] Step S262: Perform real-time correlation updates on the preprocessed data stream and the preliminary dynamic weight graph based on a sliding window to obtain the real-time correlation matrix;

[0039] Step S263: Perform dynamic weight fusion and adjustment on the preliminary dynamic weight map based on the real-time correlation matrix to obtain the intermediate dynamic weight map;

[0040] Step S264: performing physical model-based relevance correction on the intermediate dynamic weight graph to obtain a dynamic weighted graph.

[0041] The application initializes the edge weight of the initial graph structure by using the monitoring point correlation matrix, integrates the statistical relevance of historical data into the graph structure, so that the preliminary dynamic weight graph can reflect the long-term data correlation mode between monitoring points. This provides a basis for subsequent dynamic adjustment based on real-time data, and enables the model to better capture the inherent correlation between monitoring points. The real-time relevance update mechanism based on the sliding window can capture the latest data correlation changes between monitoring points and reflect them in the real-time relevance matrix. The use of the sliding window considers the timeliness of the data and avoids the influence of data fluctuations at a single time point, making the relevance analysis more stable and reliable. By dynamically fusing and adjusting the real-time relevance matrix information into the preliminary dynamic weight graph, the graph structure can reflect the latest correlation trend between monitoring points. The use of the exponential weighted average method balances the influence of historical data and real-time data, retaining the contribution of historical information while highlighting the latest data changes, making the weight adjustment smoother and more stable. The relevance correction based on the physical model integrates prior knowledge and physical laws in the field of water conservancy projects into the graph structure, making up for the possible shortcomings of purely data-driven relevance analysis. For physically strongly correlated monitoring points, even if the short-term data relevance is weak, the connection strength in the graph structure can be ensured through the correction of the physical model, improving the robustness and reliability of the model.

[0042] Preferably, step S3 comprises the following steps:

[0043] Step S31: constructing a training data set according to the dynamic topology graph to obtain a model training data set;

[0044] Step S32: obtaining water conservancy prediction demand data; constructing an ST-GCN model architecture according to the water conservancy prediction demand data to obtain a model structure configuration;

[0045] Step S33: obtaining prediction task type data; selecting a loss function and an optimizer according to the model structure configuration and the prediction task type data to obtain a training parameter configuration;

[0046] Step S34: model training and verification are performed according to the model structure configuration, the model training dataset, and the training parameter configuration, to obtain an initial ST-GCN model;

[0047] Step S35: model performance evaluation is performed on the initial ST-GCN model, to obtain a model evaluation report;

[0048] Step S36: model parameter optimization is performed on the initial ST-GCN model by using the model evaluation report, to obtain an ST-GCN model.

[0049] The application effectively combines the structural information and time series data of the water conservancy project by constructing a training dataset according to a dynamic topology graph, thereby providing a high-quality data basis for the training of the ST-GCN model. The division mode of the input and output sequences considers the prediction target and time dependence of the model, so that the model can learn the evolution law of the water conservancy project state. The division of the dataset ensures the independence of model training, verification, and testing, and can more objectively evaluate the performance of the model. The ST-GCN model architecture is constructed according to the prediction requirement data of the water conservancy project, thereby ensuring that the structure of the model matches the actual prediction task. For example, according to the number and type of the target variables to be predicted, the output layer dimension of the model is determined; according to the time granularity of the prediction, the time convolution layer parameters of the model are adjusted. The rationality of the model structure configuration directly affects the prediction accuracy and efficiency of the model. Selecting appropriate loss functions and optimizers according to the type of the prediction task can effectively guide the training process of the model and improve the prediction performance of the model. For example, for a regression task, it is common to select a mean square error loss function and an Adam optimizer. The rationality of the training parameter configuration is crucial for the convergence speed and final performance of the model. Using the constructed model structure, training dataset, and training parameter configuration to train and verify the model can enable the model to learn the spatio-temporal evolution law of the water conservancy project state, and initially obtain a usable model. The application of the early stopping strategy effectively prevents model overfitting and improves the generalization ability of the model. By evaluating the performance of the initial ST-GCN model and generating an evaluation report, the prediction ability and limitations of the model can be comprehensively understood, thereby providing a reference for subsequent model optimization. Using multiple evaluation indicators, such as RMSE, MAE, and R^2, can more objectively evaluate the performance of the model. According to the results of the model evaluation report, the model parameters are optimized, such as adjusting the learning rate, weight decay coefficient, or modifying the network structure, which can further improve the prediction accuracy and generalization ability of the model, and finally obtain a ST-GCN model with better performance, thereby providing more reliable prediction results for the safety monitoring of the water conservancy project.

[0050] Preferably, step S32 comprises the following steps:

[0051] Step S321: Spatial convolution layer configuration generation based on Chebyshev spectrum convolution is performed according to the water conservancy project prediction demand data, and the spatial convolution layer configuration is obtained;

[0052] Step S322: Time convolution layer construction is performed according to the spatial convolution layer configuration, and the time convolution layer configuration is obtained;

[0053] Step S323: Time attention mechanism integration is performed according to the spatial convolution layer configuration and the time convolution layer configuration, and the attention mechanism configuration is obtained;

[0054] Step S324: Model layering structure design is performed based on the spatial convolution layer configuration, the time convolution layer configuration and the attention mechanism configuration, and the model layering configuration is obtained;

[0055] Step S325: Activation function selection is performed on the model layering configuration, and the activation function configuration is obtained;

[0056] Step S326: Model output layer design is performed according to the model layering configuration and the activation function configuration, and the model structure configuration is obtained.

[0057] The present application can effectively process graph structure data by adopting Chebyshev spectrum convolution, and better capture the spatial dependence between monitoring points by setting the order to control the receptive field size of the convolution kernel. The output feature dimension is determined according to the predicted demand data, which ensures that the model can learn sufficient spatial feature representation, providing effective information for subsequent prediction tasks. The construction of the time convolution layer enables the model to effectively capture dynamic change patterns in time series data. By setting appropriate convolution kernel size, step and padding mode, the receptive field and output sequence length of the time convolution can be controlled, so as to better extract time features and fuse them with spatial features. The integration of the time attention mechanism enables the model to focus on the historical time that has a greater impact on the current state, thereby improving the prediction accuracy of the model. By learning the weights of different time steps, the model can better capture key information in time series data and ignore irrelevant information, improving the learning efficiency and prediction performance of the model. Combining the spatial convolution layer, time convolution layer and attention mechanism into a spatio-temporal convolution block and designing a layer-stacking structure can better learn the complex spatio-temporal dependence of the water conservancy project state. The depth and width of the layer-stacking structure and the connection mode between the blocks will affect the expression ability and learning efficiency of the model. Selecting an appropriate activation function, such as the ReLU activation function, can increase the non-linear expression ability of the model and improve the learning ability and generalization performance of the model. The selection and position of the activation function will affect the training effect and final performance of the model. Designing the output layer according to the model layer-stacking configuration and activation function configuration ensures that the output of the model matches the prediction target. For example, for a regression task, the output layer is usually a fully connected layer with the same output dimension as the number of prediction target variables, and no activation function is used. The design of the output layer directly affects the prediction results and final performance of the model.

[0058] Preferably, step S4 comprises the following steps:

[0059] Step S41: performing model input preprocessing on the original data stream and the dynamic topology graph to obtain a model input graph;

[0060] Step S42: inputting the model input graph into an ST-GCN model to perform future state prediction and obtain a predicted state sequence;

[0061] Step S43: extracting monitoring point information from the dynamic topology graph to obtain monitoring point information; performing risk preliminary judgment on the predicted state sequence by using a preset safety threshold data and the monitoring point information to obtain an individual risk assessment result;

[0062] Step S44: performing multi-parameter comprehensive risk assessment based on a Bayesian network according to the individual risk assessment result to obtain a comprehensive risk assessment result;

[0063] Step S45: risk assessment report generation is performed on the predicted state sequence, the monitoring point information, the individual risk assessment result and the comprehensive risk assessment result to obtain a risk assessment report.

[0064] Step S46: risk level determination is performed on the risk assessment report to obtain a determined risk level; according to the determined risk level and a preset early warning strategy configuration, early warning trigger condition determination is performed to obtain an early warning trigger signal.

[0065] Step S47: early warning information generation is performed according to the early warning trigger signal, the risk assessment report and the monitoring point information to obtain to-be-sent early warning information; early warning information publishing is performed on the to-be-sent early warning information to obtain a sent early warning record.

[0066] The application pre-processes the original data stream and dynamic topology map, including outlier elimination, missing value filling and data normalization, to ensure the quality and consistency of the input data, and improve the accuracy and stability of the model prediction. The pre-processed data is converted into tensor format, which facilitates efficient calculation and processing of the model. The trained ST-GCN model is used to predict the future state of the model input map, and the predicted state sequence is obtained, realizing the prediction of the future state of the water conservancy project and providing a basis for risk assessment and early warning. The prediction results are output in the form of time series, which can intuitively show the trend of the water conservancy project state in the future. Extract the monitoring point information and combine the preset safety threshold data to preliminarily judge the risk of the predicted state sequence, and obtain the individual risk assessment result of each monitoring point, which provides a basis for subsequent comprehensive risk assessment. By comparing the predicted value with the safety threshold, potential risk points can be quickly identified and their risk levels can be preliminarily assessed. The multi-parameter comprehensive risk assessment method based on Bayesian network comprehensively considers the correlation and uncertainty between multiple monitoring parameters, which can more comprehensively and accurately assess the overall risk level of the water conservancy project. The inference process of the Bayesian network can effectively handle incomplete and uncertain information, improving the reliability of risk assessment. Generate a risk assessment report, integrate the prediction results, risk assessment results and monitoring point information into a document, and facilitate relevant personnel to understand the operation state and risk status of the water conservancy project. The charts and textual descriptions included in the report make the risk information more intuitive and easy to understand, providing an important reference for decision-making. According to the risk assessment report, determine the risk level and combine the early warning strategy configuration to determine whether to trigger the early warning, realizing the automatic early warning triggering mechanism. The flexibility of the early warning strategy configuration can set different early warning triggering conditions according to different risk levels and early warning needs, improving the accuracy and efficiency of early warning. According to the early warning trigger signal, risk assessment report and monitoring point information, generate early warning information and publish it to relevant personnel, realizing timely risk early warning and information transmission. The publication of early warning information can be carried out through various channels, such as SMS, email or special APP, ensuring the timely delivery of early warning information so that relevant personnel can take timely measures to reduce risks. Record the sent early warning information to facilitate subsequent early warning management and effect evaluation.

[0067] Preferably, step S44 comprises the following steps:

[0068] Step S441: obtaining a water conservancy project safety evaluation index system and a historical risk event analysis report; defining the basic monitoring parameter node based on the individual risk assessment result, to obtain the basic monitoring parameter node;

[0069] Step S442: According to the basic monitoring parameter node and the water conservancy safety evaluation index system, an intermediate risk index node is defined, and an intermediate risk index node is obtained;

[0070] Step S443: According to the intermediate risk index node, a final risk grade node is defined, and a Bayesian network node definition table is constructed according to the basic monitoring parameter node, and the Bayesian network node definition table is obtained;

[0071] Step S444: The node directed edge is obtained by using the preset water conservancy professional knowledge base to establish the node directed edge of the Bayesian network node definition table; the causal relationship data is obtained by modeling the causal relationship of the Bayesian network node definition table;

[0072] Step S445: The network structure data is obtained by using the historical risk event analysis report to determine the network structure of the Bayesian network node definition table;

[0073] Step S446: The Bayesian network structure diagram is obtained by constructing the Bayesian network structure according to the node directed edge, the causal relationship data and the network structure data;

[0074] Step S447: The Bayesian network conditional probability table is obtained by determining the conditional probability table of the Bayesian network structure diagram;

[0075] Step S448: The individual risk assessment result is input into the Bayesian network structure diagram as evidence, and the Bayesian network conditional probability table is used for risk reasoning and evaluation, and the comprehensive risk assessment result is obtained.

[0076] The present application provides domain knowledge and data support for constructing the Bayesian network by obtaining the water conservancy project safety evaluation index system and the historical risk event analysis report, so that the Bayesian network can more accurately reflect the actual risk status of the water conservancy project. The individual risk assessment results are used to define the basic monitoring parameter nodes, the monitoring data is connected with the Bayesian network model, and the input data is provided for subsequent risk reasoning. The intermediate risk indicator nodes are defined, the risk information of the basic monitoring parameter nodes is integrated and abstracted, higher level risk indicators such as'structural deformation risk' and'seepage safety risk' are formed, and the risk assessment is more systematic and hierarchical. The final risk level nodes are defined, and the Bayesian network node definition table is constructed, the node definition of the Bayesian network model is completed, and the variables in the model and their value ranges are clearly defined, which provides a basis for subsequent construction of network structure and conditional probability table. The professional knowledge base is used to establish the node directed edge and the causal relationship model, the expert experience and domain knowledge are integrated into the Bayesian network, the network structure can more accurately reflect the causal relationship and influence mechanism between the monitoring parameters, and the scientificity and reliability of the risk assessment are improved. The historical risk event analysis report is used to determine the network structure, the correlation between the risk factors is learned from the historical data, and is reflected in the Bayesian network structure, so that the model is more in line with the actual situation, and the prediction ability of the model is improved. The Bayesian network structure diagram is constructed according to the node directed edge, the causal relationship data and the network structure data, the structure of the Bayesian network model is completed, and the dependency relationship and influence path between the risk factors are clearly shown, which provides a framework for subsequent risk reasoning. The conditional probability table is determined, the probability information is injected into the Bayesian network, the influence degree between the risk factors is quantified, and the risk assessment is more refined and quantified. The conditional probability table can be determined based on historical data statistics, expert experience or machine learning method, so as to ensure the accuracy and reliability of the probability value. The individual risk assessment results are input into the Bayesian network as evidence, and the conditional probability table is used for risk reasoning, and the comprehensive risk assessment result is obtained, and the overall risk of the water conservancy project is comprehensively evaluated. The reasoning process of the Bayesian network can effectively handle uncertain information, and the accuracy and reliability of the risk assessment are improved. BRIEF DESCRIPTION OF DRAWINGS

[0077] Fig. 1 It is a step flowchart of a kind of water conservancy project safety monitoring method based on data processing;

[0078] Fig. 2 It is a detailed implementation step flowchart of step S2 in the present application.

[0079] The object of the present application, functional characteristics and advantages will be further described with reference to the embodiments and accompanying drawings. DETAILED DESCRIPTION

[0080] The technical method of the present application will be described clearly and completely below in conjunction with the drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0081] In addition, the drawings are only schematic illustrations of the present application and are not necessarily drawn to scale. Identical reference numerals in the drawings represent identical or similar parts, and thus repeated descriptions thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities, which do not necessarily have to correspond to physically or logically independent entities. The functional entities can be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0082] It should be understood that although the terms "first", "second" and the like can be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, without departing from the scope of the exemplary embodiments, a first element can be referred to as a second element, and similarly a second element can be referred to as a first element. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0083] To achieve the above-mentioned purposes, please refer to Figs. 1-2 A water conservancy project safety monitoring method based on data processing, comprising the following steps:

[0084] Step S1: Constructing a multi-source data acquisition scheme for the water conservancy project to obtain a multi-source data acquisition scheme; collecting multi-source data for the water conservancy project according to the multi-source data acquisition scheme to obtain an original data stream;

[0085] Step S2: Hierarchical monitoring nodes are constructed according to the multi-source data acquisition scheme to obtain an initial node set; initial edges are constructed and defined according to the initial node set to obtain an initial edge list; different types of initial edge weights are assigned to the initial edge list, and an initial graph structure is generated to obtain an initial graph structure; correlation analysis is performed on the original data stream to obtain a monitoring point correlation matrix; the initial graph structure is adjusted and processed by using the monitoring point correlation matrix to obtain a dynamic topology graph;

[0086] Step S3: obtaining water conservancy project predicted demand data; constructing an ST-GCN model architecture according to the water conservancy project predicted demand data to obtain a model structure configuration; constructing an ST-GCN model according to the model structure configuration and a dynamic topology atlas, and performing ST-GCN model training;

[0087] Step S4: predicting a future state using the ST-GCN model to obtain a predicted state sequence; performing a risk preliminary judgment on the predicted state sequence to obtain an individual risk evaluation result; performing a multi-parameter comprehensive risk evaluation based on a Bayesian network according to the individual risk evaluation result to obtain a comprehensive risk evaluation result; generating a risk evaluation report according to the comprehensive risk evaluation result to obtain a risk evaluation report; judging a warning triggering condition for the risk evaluation report and publishing warning information to realize a water conservancy project safety monitoring task.

[0088] In the embodiment of the present application, as shown in the reference Fig. 1 The data processing-based water conservancy project safety monitoring method includes the following steps:

[0089] Step S1: constructing a multi-source data acquisition scheme for a water conservancy project to obtain a multi-source data acquisition scheme; acquiring multi-source data of the water conservancy project according to the multi-source data acquisition scheme to obtain an original data stream;

[0090] In the embodiment of the present application, the core is to collect data from various sensors on the dam and perform preprocessing to make it available for subsequent analysis. Mainly includes: determining the sensor type and position according to the dam structure and monitoring requirements, establishing a data transmission network (for example, using LoRaWAN), collecting real-time data stream, converting the original sensor data into physical quantities (such as displacement, pressure), verifying data integrity through range check and outlier detection (such as 3σ criterion), and finally synchronizing and integrating multi-source data into a unified, time-stamped data stream.

[0091] Specifically: deploying Trimble NetR9 GNSS receivers on the dam crest and Geokon 4500S-ATM vibrating wire piezometers inside the dam body, transmitting data to the central server through the LoRaWAN network. Collecting data at a preset frequency (for example, GNSS 1Hz, piezometer 1 / 60Hz). Convert raw data to physical units using sensor-specific calibration parameters. Data verification includes range check and 3σ outlier detection. Use linear interpolation for time synchronization, and integrate the verified data into a unified CSV file containing timestamps, sensor IDs and measurement values.

[0092] Step S2: hierarchical monitoring node is obtained according to a multi-source data acquisition scheme; an initial edge list is obtained according to the initial node set; different type initial edge weight assignment is performed on the initial edge list, and initial graph structure generation is performed to obtain the initial graph structure; correlation analysis is performed on the original data stream to obtain a monitoring point correlation matrix; and the initial graph structure is processed by using the monitoring point correlation matrix to adjust the dynamic edge weight, and a dynamic topology graph is obtained;

[0093] In the embodiment of the application, the target is to create a dynamic graph representing the dam and its sensor network. Mainly includes: mapping the sensors to the locations on the dam CAD model, assigning importance levels according to the locations and functions of the monitoring points, defining edges based on physical connections and spatial proximity (for example, a 5-meter threshold), assigning initial edge weights according to material properties (elastic modulus of physical connection, inverse distance of spatial proximity), calculating dynamic edge weights based on real-time and historical data correlation (using Pearson correlation coefficient with a 7-day sliding window), and finally fusing the dynamic graph structure with the preprocessed sensor data to create a dynamic topology graph.

[0094] Specifically: using the dam CAD model in AutoCAD, mapping the sensor ID to its 3D coordinates, and assigning importance levels (1, 2 or 3). Create edges between nodes that are physically connected and within a 5-meter range, and assign initial weights of 1.0 and inverse distance, respectively. Calculate the Pearson correlation coefficient between the time series of sensor data using a 7-day sliding window. Update the edge weights according to these correlations (using a threshold of 0.8 and an update coefficient of 0.7). Combine the dynamic graph structure with the preprocessed sensor data from S1, and store the graph and node features at each time step in a custom binary file format.

[0095] Step S3: obtain water conservancy project prediction demand data; construct an ST-GCN model architecture according to the water conservancy project prediction demand data to obtain a model structure configuration; construct an ST-GCN model according to the model structure configuration and the dynamic topology graph, and perform ST-GCN model;

[0096] In the embodiment of the application, the focus is on training the spatio-temporal graph convolutional network (ST-GCN) model. Mainly includes: preparing a training data set by creating input-output sequences from the dynamic topology graph (for example, T = 24-hour input, P = 1-hour prediction range), defining the ST-GCN architecture (using Chebyshev graph convolution and one-dimensional time convolution), selecting the loss function (MSE for regression) and the optimizer (Adam), training the model using TensorFlow and using the early stopping strategy to prevent overfitting, evaluating the performance of the model on the test set using RMSE, MAE and R^2, and finally optimizing the model parameters according to the evaluation results.

[0097] Specifically: Prepare the training dataset by creating 24-hour input and 1-hour output sequences from the dynamic topology graph. Define an ST-GCN model containing two blocks, including Chebyshev graph convolution (2nd order, 64 output features), one-dimensional temporal convolution (kernel size 3, 64 output features), and multi-head attention mechanism (8 heads, dimension 8). Use the MSE loss and Adam optimizer (learning rate 0.001, weight decay 0.0001). Train the model using TensorFlow and early stop according to the validation loss. Evaluate the model on the test set using RMSE, MAE, and R^2. If the performance is below the expected threshold, optimize the model parameters and architecture.

[0098] Step S4: Perform future state prediction using the ST-GCN model to obtain a predicted state sequence; perform a preliminary risk assessment on the predicted state sequence to obtain an individual risk assessment result; perform a multi-parameter comprehensive risk assessment based on a Bayesian network according to the individual risk assessment result to obtain a comprehensive risk assessment result; generate a risk assessment report according to the comprehensive risk assessment result to obtain a risk assessment report; determine a pre-warning trigger condition for the risk assessment report and publish pre-warning information to achieve the water conservancy project safety monitoring task;

[0099] In the embodiment of the application, the trained ST-GCN model is used for prediction and risk assessment, real-time data is preprocessed and the dynamic topology graph is updated, the ST-GCN model is used to predict the future state (e.g. 1 hour in advance), the predicted value is compared with the predefined safety threshold, the Bayesian network is used (combined with expert knowledge and historical data) to perform multi-parameter risk assessment, a risk assessment report is generated, the risk level is determined according to the Bayesian network output, the alarm is triggered according to the predefined rules and the evaluated risk level (different alarm levels are set for different risk severity), and finally the alarm is sent to the designated recipient through SMS, email or a dedicated application.

[0100] Specifically: preprocess the real-time data and update the dynamic topology graph as in step S1. Use the trained ST-GCN model to predict the state 1 hour in the future. Compare the predicted value with the predefined threshold for each monitoring parameter. Input the individual risk assessment into the pre-constructed Bayesian network, use belief propagation to infer the overall risk level, and generate a risk assessment report (PDF). Determine the overall risk level. Trigger the alarm according to the risk level and predefined rules (e.g. "alert" level immediately issues an alarm, "warning" level issues an alarm after more than 1 hour).

[0101] Preferably, step S1 includes the following steps:

[0102] Step S11: Sensor deployment for water conservancy projects and data transmission network construction to obtain a multi-source data collection scheme;

[0103] Step S12: Multi-source data collection for water conservancy projects according to the multi-source data collection scheme to obtain real-time multi-source data flow;

[0104] Step S13: Data format conversion and verification of real-time multi-source data flow to obtain formatted and verified data;

[0105] Step S14: Multi-source data integration and time synchronization of formatted and verified data to obtain an original data stream.

[0106] In the embodiment of the application, first, according to the dam design drawings and the predetermined monitoring requirements, the key parameters that need to be monitored are determined, such as dam deformation, seepage pressure, temperature, water level, etc. Then, select the appropriate sensor type for each monitoring parameter, such as GNSS receiver for monitoring dam deformation, vibrating string seepage pressure gauge for monitoring seepage pressure, PT100 temperature sensor for monitoring temperature, and ultrasonic water level gauge for monitoring water level. Each sensor type must be specified, such as TrimbleNetR9 for GNSS receiver and Geokon4500S-ATM for vibrating string seepage pressure gauge. Next, according to the dam structure and the accessibility of the monitoring points, the specific installation location and installation method of each sensor are determined. For example, the GNSS receiver is installed on the control point on the dam top with bolt fixation, and the seepage pressure gauge is buried at a predetermined depth inside the dam body and connected with the data collector. All installation locations and fixation methods must be recorded in detail in the multi-source data collection scheme. Finally, the data transmission network is constructed. LoRa wireless communication technology is adopted, LoRaWAN gateway is deployed in the dam area, and each sensor is connected to the LoRaWAN node. The parameters of the LoRaWAN node, such as frequency, spreading factor, and coding rate, must be configured according to the actual situation to ensure reliable data transmission. The multi-source data collection scheme contains all the specific configuration information of sensor types, installation locations, installation methods, and data transmission network.

[0107] According to the multi-source data acquisition scheme determined in step S11, start the installed sensors to collect data. Each sensor collects data according to the pre-set sampling frequency, for example, the GNSS receiver collects data once every second, and the osmometer collects data once every minute. The sampling frequency of the sensor must be clearly defined in the multi-source data acquisition scheme. The collected data is transmitted to the data center server in real time through the LoRaWAN network. The data center server continuously receives data packets from each LoRaWAN node and parses the sensor ID, timestamp and original measurement value contained in each data packet. These data constitute a real-time multi-source data stream in the form of time series, each data point containing sensor ID, timestamp and original measurement value. The format of the data stream must be pre-defined, such as using JSON format or CSV format.

[0108] The real-time multi-source data stream obtained in step S12 is converted and checked. First, according to the communication protocol and data format specification of each sensor, the original measurement value is converted into the corresponding physical quantity, for example, the original data of the GNSS receiver is converted into latitude and longitude coordinates and elevation value, and the frequency data of the osmometer is converted into pressure value. The conversion formula must be clearly listed, for example, pressure value = a*frequency^2 + b*frequency + c, where a, b, c are calibration coefficients. Then, the converted data is checked. The checking rules must be pre-set, for example, check whether the data is within a reasonable range, whether there are obvious mutations or abnormal values. 3σ criterion or other statistical methods can be used for abnormal value detection. If the data fails the check, record the error information and mark the data point as invalid. The data that passes the check is added with a check mark and constitutes the formatted and checked data together with the sensor ID and timestamp.

[0109] The formatted and checked data obtained in step S13 is integrated and time-synchronized to generate the final raw data stream. First, according to the timestamp, align the data from different sensors. Since the sampling frequencies of different sensors may be different, time synchronization processing is needed. Linear interpolation method is used to interpolate the data of sensors with lower sampling frequency to the same time interval as the sensor with the highest sampling frequency. The interpolation method and parameters must be clearly set, for example, linear interpolation is used and the time interval is 1 second. Then, sort all sensor data by timestamp and integrate them into a unified data structure. The data structure must be pre-defined, for example, in the form of a table containing timestamp, sensor ID and corresponding measurement value. Finally, check the quality of the integrated data, for example, check whether there are data missing or duplication. The checking method must be clearly defined, for example, check whether there is data of all sensors under each timestamp. The data after integration and time synchronization constitutes the final raw data stream, which is used for subsequent analysis and processing.

[0110] Preferably, step S2 comprises the following steps:

[0111] Step S21: Obtain a water conservancy structure design drawing; perform monitoring node mapping on the multi-source data collection scheme and the water conservancy structure design drawing to obtain a node mapping set;

[0112] Step S22: Perform node importance classification on the node mapping set to obtain an initial node set;

[0113] Step S23: Perform initial edge construction and type definition according to the water conservancy structure design drawing and the initial node set to obtain an initial edge list; perform different type initial edge weight assignment on the initial edge list to obtain a weighted initial edge list;

[0114] Step S24: Perform initial graph structure generation according to the weighted initial edge list and the initial edge list to obtain an initial graph structure;

[0115] Step S25: Perform correlation analysis on the original data stream based on historical data to obtain a monitoring point correlation matrix; perform original data preprocessing on the original data stream to obtain a preprocessed data stream;

[0116] Step S26: Perform dynamic edge weight adjustment on the initial graph structure by using the monitoring point correlation matrix to obtain a dynamic weighted graph;

[0117] Step S27: Perform data fusion on the dynamic weighted graph and the preprocessed data stream to obtain a dynamic topology graph.

[0118] As an example of the present application, reference is made to Fig. 1, which shows a flowchart of a method for constructing a dynamic topology graph of a water conservancy structure according to an embodiment of the present application. In this example, the step S2 comprises: Fig. 2

[0119] Step S21: Obtain a water conservancy structure design drawing; perform monitoring node mapping on the multi-source data collection scheme and the water conservancy structure design drawing to obtain a node mapping set;

[0120] ​In the embodiment of the application, the CAD file of the water conservancy project structure design drawing is acquired, which contains the geometric shape of the dam, coordinate information of key parts and the like. The CAD file is opened by using AutoCAD software, and the three-dimensional coordinate information of all monitoring points on the dam body is extracted. The coordinate accuracy of the monitoring points must reach the millimeter level. Then, according to the multi-source data acquisition scheme generated in step S11, the ID of each sensor in the scheme is mapped to the corresponding monitoring point on the structure design drawing one by one. For example, the GNSS receiver with ID GNSS_001 is mapped to the dam crest control point A, and the seepage pressure gauge with ID SP_001 is mapped to the monitoring point B inside the dam body. Each mapping relationship must be recorded clearly, and a node mapping set containing the sensor ID, monitoring point name and three-dimensional coordinates is generated. The node mapping set is stored in the CSV file format, which contains three columns: sensor ID, monitoring point name, X coordinate, Y coordinate and Z coordinate.

[0121] Step S22: node importance classification is performed on the node mapping set to obtain an initial node set;

[0122] In the embodiment of the application, based on the water conservancy project safety monitoring specification and expert experience, the importance of each monitoring point in the node mapping set generated in step S21 is classified. The classification standard is based on the importance of the monitoring point position and the engineering safety state reflected thereby. For example, the monitoring points located at the main section of the dam body and the foundation part are classified as first-level nodes, with the highest importance; the monitoring points located at the dam shoulder and non-key parts are classified as second-level nodes; and the monitoring points located at the surrounding environment monitoring are classified as third-level nodes, with the lowest importance. The importance level of each node is represented by the numbers 1, 2 and 3, and is added to the node mapping set to form an initial node set. The initial node set is stored as a CSV file, which contains four columns: sensor ID, monitoring point name, three-dimensional coordinates and importance level.

[0123] Step S23: initial edge construction and type definition are performed according to the water conservancy project structure design drawing and the initial node set to obtain an initial edge list; and different type initial edge weight values are assigned to the initial edge list to obtain a weighted initial edge list;

[0124] In the embodiment of the present application, an initial edge list is constructed according to the water conservancy project structure design drawing and the initial node set. The connection rules are as follows: for any two monitoring points, if they are directly connected in physics (for example, located in the same dam section), a "physical connection" edge is established between them; if the spatial distance between them is less than 10 meters, a "spatial proximity" edge is established. The type of the edge must be clearly defined. The initial edge list is stored as a CSV file, containing three columns: starting node ID, ending node ID, and edge type. Then, different types of initial edge weight assignment are performed on the initial edge list. The initial weight of the physical connection edge is set to 1.0, and the initial weight of the spatial proximity edge is set to the inverse of the distance between the nodes. The weight value must be explicitly set. The weight value is added to the initial edge list to form a weighted initial edge list, which is stored as a CSV file, containing four columns: starting node ID, ending node ID, edge type, and weight.

[0125] Step S24: generating an initial graph structure according to the weighted initial edge list and the initial edge list;

[0126] In the embodiment of the present application, the NetworkX graph processing library is used to construct the initial graph structure based on the weighted initial edge list generated in step S23. Each monitoring point in the initial node set is taken as a node of the graph, and each edge in the weighted initial edge list is taken as an edge of the graph. The node attributes include the monitoring point name, three-dimensional coordinates, and importance level. The edge attributes include the edge type and weight. The generated initial graph structure is saved in GraphML file format, containing node information, edge information, and their attributes.

[0127] Step S25: performing historical data-based correlation analysis on the original data stream to obtain a monitoring point correlation matrix; and performing original data preprocessing on the original data stream to obtain a preprocessed data stream;

[0128] In the embodiment of the present application, historical data-based correlation analysis is performed on the original data stream obtained in step S14. Past one year of historical data is selected, and the Pearson correlation coefficient of the time series data between any two monitoring points is calculated. The calculation method of the Pearson correlation coefficient must be clearly defined. The calculation results are stored in an N x N matrix, where N is the number of monitoring points, and each element of the matrix represents the correlation coefficient between the corresponding two monitoring points, forming a monitoring point correlation matrix. The matrix is stored in CSV file format. Then, the original data stream is preprocessed. Abnormal values are removed using the 3σ criterion, and missing values are filled using the linear interpolation method. The preprocessing method and parameters must be explicitly set. The data is normalized by scaling the value of each monitoring parameter to the [0, 1] interval. The normalization method must be clearly defined, for example, using Min-Max normalization. The preprocessed data is stored as a CSV file, forming a preprocessed data stream.

[0129] Step S26: dynamically adjusting the initial graph structure by using the monitoring point correlation matrix to obtain a dynamically weighted graph;

[0130] In the embodiment of the application, the initial graph structure generated in step S24 is dynamically adjusted in edge weight based on the monitoring point correlation matrix generated in step S25. For each edge in the initial graph structure, if the corresponding element value (i.e., the correlation coefficient) of the two monitoring points connected in the monitoring point correlation matrix is greater than 0.8, the weight of the edge is updated to the correlation coefficient; otherwise, the initial weight of the edge is kept unchanged. The weight adjustment rule must be clearly defined. The adjusted graph structure is referred to as a dynamically weighted graph, which is still saved in the GraphML file format and contains updated edge weight information.

[0131] Step S27: data fusion is performed on the dynamically weighted graph and the preprocessed data stream to obtain a dynamic topology graph;

[0132] In the embodiment of the application, the dynamically weighted graph generated in step S26 and the preprocessed data stream generated in step S25 are fused to generate a dynamic topology graph. The monitoring data of each time step in the preprocessed data stream is taken as a feature vector of the corresponding node. The dynamic topology graph contains the graph structure (node, edge, weight) of each time step and the feature vector of the node. The dynamic topology graph is stored in a self-defined binary file format, and the data of one graph is stored for each time step, including node information, edge information, weight information and node feature vectors.

[0133] Preferably, step S23 comprises the following steps:

[0134] Step S231: marking physical connection edges for the initial node set to obtain physical connection edges; and marking spatial proximity edges for the initial node set to obtain spatial proximity edges;

[0135] Step S232: marking water flow direction associated edges for the initial node set to obtain water flow associated edges; and marking force transmission path associated edges for the initial node set to obtain force associated edges;

[0136] Step S233: constructing an initial edge list according to the physical connection edges, the spatial proximity edges, the water flow associated edges and the force associated edges to obtain the initial edge list;

[0137] Step S234: assigning weights to the physical connection edges based on a material strength parameter to obtain physical connection edge weights; and assigning weights to the spatial proximity edges based on the reciprocal of the Euclidean distance between nodes to obtain spatial proximity edge weights;

[0138] Step S235: weight assignment based on fixed weight value of directed edge is performed on the water flow associated edge to obtain water flow direction associated edge weight; weight assignment based on stress transmission coefficient is performed on the force associated edge to obtain force transmission path associated edge weight;

[0139] Step S236: weighted initial edge list construction is performed according to the physical connection edge weight, the spatial adjacent edge weight, the water flow direction associated edge weight and the force transmission path associated edge weight to obtain the weighted initial edge list.

[0140] In the embodiment of the application, the initial node set CSV file is read to obtain the ID and three-dimensional coordinates of each monitoring point. Then, the physical connection relationship is determined according to the water conservancy engineering structure design drawing. For example, for an arch dam, there is a physical connection between adjacent monitoring points; for a gravity dam, there is a physical connection between monitoring points in the same dam section. The node pairs with physical connection are recorded and marked as “physical connection” edges, and stored as a CSV file containing two columns of starting node ID and ending node ID. Then, the Euclidean distance between any two nodes in the initial node set is calculated. The distance is calculated using three-dimensional coordinates, and the formula is `distance=sqrt((x1-x2)^2+(y1-y2)^2+(z1-z2)^2)`. If the distance between the two nodes is less than or equal to 5 meters, a “spatial adjacent” edge is established between the two nodes. The spatial adjacent edge is also saved as a CSV file containing two columns of starting node ID and ending node ID.

[0141] The initial node set CSV file is read to obtain the ID, position information and monitoring parameter type of each monitoring point. The water flow direction associated edge is determined according to the water conservancy engineering design drawing and the water flow direction. For example, for seepage monitoring points, according to the flow direction of water flow from upstream to downstream, a directed edge is established from the upstream monitoring point to the downstream monitoring point. These edges are marked as “water flow associated” edges and stored as a CSV file containing two columns of starting node ID and ending node ID. Then, the force transmission path associated edge is determined according to the finite element analysis result or other stress analysis method. For example, for stress monitoring points, according to the stress transmission path, a directed edge is established from the force applying point to the force receiving point. These edges are marked as “force associated” edges and also stored as a CSV file containing two columns of starting node ID and ending node ID.

[0142] The four CSV files generated in reading steps S231 and S232: physical connection edges, spatial proximity edges, water flow correlation edges, and force correlation edges. Merge the edge information in the four files into a new CSV file to form an initial edge list. The CSV file contains three columns: starting node ID, ending node ID, and edge type. The edge type field indicates the type of the edge, such as "physical connection", "spatial proximity", "water flow correlation" or "force correlation". Ensure that there are no duplicate edges in the merging process, and if there are duplicate edges, keep one and record the duplicate information.

[0143] Read the initial edge list generated in step S233 and the water conservancy material parameter table. For each "physical connection" edge, find the corresponding elastic modulus from the material parameter table according to the material type and strength grade of the connection site. Set the weight of the edge to the elastic modulus value. For example, if the material of the connection site is C30 concrete, the weight of the edge is set to 3.25x10^10 Pa. For each "spatial proximity" edge, calculate the Euclidean distance `d` between the two nodes, and set the weight of the edge to `1 / d`. The calculation result is kept to four significant digits. Store the calculated weight values into two new CSV files respectively, corresponding to the physical connection edge weight and the spatial proximity edge weight, each file contains three columns: starting node ID, ending node ID and weight.

[0144] Read the initial edge list generated in step S233 and the finite element analysis result file. For each "water flow correlation" edge, assign a fixed weight value of 0.5 to represent the influence of the water flow direction. Save the results to a CSV file containing three columns: starting node ID, ending node ID and weight. For each "force correlation" edge, extract the stress transfer coefficient connecting the two nodes from the finite element analysis result file. Set the weight of the edge to the absolute value of the stress transfer coefficient. Save the results to another CSV file containing three columns: starting node ID, ending node ID and weight.

[0145] Read the four CSV files generated in steps S234 and S235: physical connection edge weight, spatial proximity edge weight, water flow correlation edge weight, and force correlation edge weight. Create a new CSV file named weighted initial edge list. Merge all edge information in the four files into the weighted initial edge list. The file contains four columns: starting node ID, ending node ID, edge type and weight. Ensure that the weight value of each edge matches its corresponding edge type.

[0146] Preferably, step S26 comprises the following steps:

[0147] Step S261: using the monitoring point correlation matrix to perform statistical correlation-based weight initialization on the initial graph structure to obtain a preliminary dynamic weight graph;

[0148] Step S262: Real-time correlation update based on a sliding window is performed on the preprocessed data stream and the preliminary dynamic weight graph to obtain a real-time correlation matrix.

[0149] Step S263: Dynamic weight fusion and adjustment are performed on the preliminary dynamic weight graph according to the real-time correlation matrix to obtain an intermediate dynamic weight graph.

[0150] Step S264: Correlation correction based on a physical model is performed on the intermediate dynamic weight graph to obtain a dynamic weighted graph.

[0151] In the embodiment of the application, the initial graph structure (in GraphML format) generated in step S24 and the monitoring point correlation matrix (in CSV format) generated in step S25 are read. A new graph structure is created, and all nodes and edge attributes of the initial graph structure are copied. For each edge in the graph, the IDs of the two connected nodes (for example, Node_A and Node_B) are obtained. The Pearson correlation coefficient corresponding to the two nodes in the monitoring point correlation matrix is found (for example, `correlation(Node_A,Node_B)`). The absolute value of the correlation coefficient is taken as the initial dynamic weight of the edge. If there is no corresponding correlation coefficient in the monitoring point correlation matrix for the two nodes (for example, due to data missing), the initial dynamic weight of the edge is set to the weight of the edge in the initial graph structure. The updated edge weight information is saved to the new graph structure, and the graph structure is saved in GraphML format, named as the preliminary dynamic weight graph.

[0152] The preprocessed data stream (in CSV format) generated in step S25 and the preliminary dynamic weight graph (in GraphML format) generated in step S261 are read. The size of the sliding window is set to 7 days, i.e. the data of the last 168 hours. For the current time *t*, the data of 168 hours before *t* in the preprocessed data stream is extracted. For each edge in the preliminary dynamic weight graph, the IDs of the two connected nodes are obtained. The Pearson correlation coefficient of the data sequence of the two nodes within the sliding window is calculated. The calculated correlation coefficient is stored in an N×N matrix, where N is the number of monitoring points, forming a real-time correlation matrix. If the data of the two nodes within the sliding window is insufficient to calculate the correlation coefficient, the weight of the corresponding edge in the preliminary dynamic weight graph is used as a substitute value for the correlation coefficient. The real-time correlation matrix is saved in CSV format.

[0153] The preliminary dynamic weight graph (in GraphML format) generated in step S261 and the real-time correlation matrix (in CSV format) generated in step S262 are read. The weight update coefficient a is set to 0.7. A new graph structure is created, copying all nodes and edge attributes of the preliminary dynamic weight graph. For each edge in the graph, the IDs of the two connected nodes (e.g., Node_A and Node_B) are obtained. The real-time correlation coefficient of the two nodes is obtained from the real-time correlation matrix (e.g., `realtime_correlation(Node_A,Node_B)`). The weight of the edge is updated using the exponential weighted average method: `W_new(Node_A,Node_B) = a * |realtime_correlation(Node_A,Node_B)| + (1-a) * W_old(Node_A,Node_B)`, where `W_old(Node_A,Node_B)` is the weight of the edge in the preliminary dynamic weight graph. The updated weight is limited to the range [0, 1]. The updated edge weight information is saved to the new graph structure, and the graph structure is saved in GraphML format, named as the intermediate dynamic weight graph atlas.

[0154] The intermediate dynamic weight graph atlas (in GraphML format) generated in step S263 and the pre-established physical correlation knowledge base are read. The physical correlation knowledge base is a CSV file containing three columns: starting node ID, ending node ID, and correlation strength. The correlation strength is a numerical value representing the strength of the physical correlation between two nodes, for example, 1 represents strong correlation, 0.5 represents moderate correlation, and 0 represents no correlation. A new graph structure is created, copying all nodes and edge attributes of the intermediate dynamic weight graph atlas. For each edge in the graph, the IDs of the two connected nodes are obtained. The correlation strength of the two nodes is looked up in the physical correlation knowledge base. If the correlation strength is greater than 0.8 and the weight of the corresponding edge in the intermediate dynamic weight graph atlas is less than 0.5, the weight of the edge is updated to 0.5. This ensures that even if the short-term data correlation is low, the edge weight between physically strongly correlated nodes remains at a reasonable level. The updated edge weight information is saved to the new graph structure, and the graph structure is saved in GraphML format, named as the dynamically weighted graph.

[0155] Preferably, step S3 comprises the following steps:

[0156] Step S31: constructing a training data set according to the dynamic topology atlas, to obtain a model training data set;

[0157] Step S32: obtaining water conservancy project prediction demand data; constructing an ST-GCN model architecture according to the water conservancy project prediction demand data, to obtain a model structure configuration;

[0158] Step S33: Obtain prediction task type data; according to the model structure configuration and the prediction task type data, loss function and optimizer selection are performed to obtain training parameter configuration;

[0159] Step S34: According to the model structure configuration, the model training data set and the training parameter configuration, the model training and verification are performed to obtain the initial ST-GCN model;

[0160] Step S35: The model performance evaluation is performed on the initial ST-GCN model to obtain a model evaluation report;

[0161] Step S36: The model parameter optimization is performed on the initial ST-GCN model by using the model evaluation report to obtain the ST-GCN model.

[0162] In the embodiment of the application, the dynamic topology atlas data generated in the reading step S27 is read. The data contains the graph structure (nodes, edges, weights) and node features (sensor monitoring data) of each time step. The input time step length is set to T=24 (i.e. the data of the past 24 hours), and the prediction time step length is set to P=1 (i.e. the state of the future 1 hour is predicted). Continuous time series data is extracted from the dynamic topology atlas to construct training samples. Each training sample contains an input sequence and an output sequence. The input sequence is a sequence composed of graph data of the past T time steps, and the output sequence is the node features of the future P time steps. The data set is divided into a training set, a validation set and a test set according to a ratio of 7:2:1. The training set, the validation set and the test set are saved as three independent binary files to constitute the model training data set.

[0163] The water conservancy project prediction requirement data is obtained, which clearly indicates the target variables to be predicted, such as dam displacement, seepage pressure, etc., and the time granularity of prediction, such as hourly, daily. According to the prediction requirement data, the ST-GCN model architecture is constructed. The model includes two spatio-temporal convolution blocks, each block consisting of a spatial graph convolution layer and a temporal convolution layer. The spatial graph convolution layer adopts Chebyshev graph convolution with order 2. The temporal convolution layer adopts one-dimensional convolution with kernel size 3. The model also includes an output layer for outputting the prediction result. The specific parameters of the model architecture, including the convolution kernel size, the feature dimension, the number of layers, etc., are recorded in the model structure configuration file and stored in JSON format.

[0164] Obtain prediction task type data, determine whether the prediction task is a regression task or a classification task. Since the prediction in the present application is continuous numerical data (such as displacement, pressure), the prediction task type is a regression task. According to the model structure configuration and the regression task type, the mean square error (MSE) is selected as the loss function, and the Adam optimizer is selected as the optimizer for model training. The learning rate of the Adam optimizer is set to 0.001, and the weight decay coefficient is set to 0.0001. The selected loss function, optimizer, and corresponding hyperparameters are recorded in the training parameter configuration file, stored in JSON format.

[0165] Using the deep learning framework TensorFlow, the ST-GCN model is constructed according to the model structure configuration generated in step S32. The model training data set generated in step S31 is loaded, and the loss function and optimizer are set according to the training parameter configuration generated in step S33. The model is trained using the training set data, and the model performance is evaluated using the validation set data after each epoch. During training, the changes in training loss and validation loss are monitored, and the early stopping strategy is used to prevent overfitting. For example, if the validation loss does not decrease for 5 consecutive epochs, the training is stopped. The model with the best performance during training is saved as the initial ST-GCN model.

[0166] Load the initial ST-GCN model obtained in step S34 and the test set data generated in step S31. Evaluate the model using the test set data, calculate the root mean square error (RMSE), mean absolute error (MAE), and determination coefficient (R^2) of the model on the test set. Record the evaluation results in the model evaluation report, stored in text file format.

[0167] Analyze the model evaluation report generated in step S35. If the model performance does not meet the pre-set requirements, such as RMSE greater than the pre-set threshold, the model parameters need to be optimized. The model hyperparameters can be adjusted, such as learning rate, weight decay coefficient, convolution kernel size, etc., or the model structure can be modified, such as increasing or decreasing the number of network layers. Repeat steps S34 and S35 until the model performance meets the requirements. The final model is saved as an ST-GCN model file.

[0168] Preferably, step S32 comprises the following steps:

[0169] Step S321: Generate a spatial convolution layer configuration based on Chebyshev spectrum convolution according to the water conservancy project prediction requirement data, to obtain a spatial convolution layer configuration;

[0170] Step S322: Construct a time convolution layer according to the spatial convolution layer configuration, to obtain a time convolution layer configuration;

[0171] Step S323: integrating the time attention mechanism according to the spatial convolution layer configuration and the time convolution layer configuration to obtain an attention mechanism configuration;

[0172] Step S324: designing a model layer structure based on the spatial convolution layer configuration, the time convolution layer configuration and the attention mechanism configuration to obtain a model layer configuration;

[0173] Step S325: selecting an activation function for the model layer configuration to obtain an activation function configuration;

[0174] Step S326: designing a model output layer according to the model layer configuration and the activation function configuration to obtain a model structure configuration.

[0175] In the embodiment of the application, the water conservancy project prediction demand data is read to determine the number and type of prediction target variables. For example, the prediction target variables are dam displacement and seepage pressure, and there are 2 variables. Based on this, the spatial convolution layer is configured. Chebyshev spectral convolution is used as the convolution operator. The order of the Chebyshev polynomial is set to 2 to capture local graph structure information. The output feature dimension of the spatial convolution layer is set to 64. The convolution kernel parameters are initialized using the Glorot uniform distribution initialization method. These configuration parameters, including the convolution operator type, order, output feature dimension and initialization method, are saved to the spatial convolution layer configuration file (JSON format).

[0176] The spatial convolution layer configuration generated in step S321 is read. According to the configuration, the time convolution layer is constructed. One-dimensional convolution is used as the time convolution operator. The convolution kernel size is set to 3, the step is set to 1, and the padding mode is set to "same" to keep the time series length unchanged. The output feature dimension of the time convolution layer is set to be the same as the output feature dimension of the spatial convolution layer, that is, 64. The convolution kernel parameters are initialized using the Glorot uniform distribution initialization method. These configuration parameters, including the convolution operator type, convolution kernel size, step, padding mode, output feature dimension and initialization method, are saved to the time convolution layer configuration file (JSON format).

[0177] The spatial convolution layer configuration generated in step S321 and the temporal convolution layer configuration generated in step S322 are read. Based on these configurations, the integrated temporal attention mechanism is designed. Using the multi-head self-attention mechanism, the number of attention heads is set to 8. The dimensions of the query, key, and value of each attention head are set to the output feature dimension of the spatial convolution layer and the temporal convolution layer divided by the number of attention heads, i.e., 64 / 8 = 8. The scaled dot-product attention calculation formula is used: Attention(Q, K, V) = softmax(QK^T / sqrt(d_k))V, where d_k is the dimension of the key. These configuration parameters, including the attention mechanism type, the number of attention heads, the query / key / value dimensions, and the calculation formula, are saved to the attention mechanism configuration file (JSON format).

[0178] The configurations generated in steps S321, S322, and S323 are read. The model layer stack structure is designed. The model contains two spatio-temporal convolution blocks. Each spatio-temporal convolution block is composed of a spatial convolution layer, a temporal convolution layer, and a temporal attention mechanism, connected in this order. A batch normalization layer and a ReLU activation function are added after each spatio-temporal convolution block. The two spatio-temporal convolution blocks are stacked together. The layer stack information, including the number of blocks, the layer type and connection order within each block, and the connection method between blocks, is saved to the model layer stack configuration file (JSON format).

[0179] The model layer stack configuration generated in step S324 is read. The activation function is selected for each spatio-temporal convolution block. After the batch normalization layer of each spatio-temporal convolution block, the ReLU activation function is used to increase the nonlinear representation ability of the model. The selection and location information of the activation function is saved to the activation function configuration file (JSON format).

[0180] The model layer stack configuration generated in step S324 and the activation function configuration generated in step S325 are read. The model output layer is designed. A fully connected layer is added at the end of the model, with an output dimension equal to the number of predicted target variables (e.g., dam displacement and seepage pressure, with an output dimension of 2). Since the prediction target is a continuous value, the output layer does not use an activation function. The type of output layer, output dimension, and activation function information are added to the model layer stack configuration and activation function configuration to form a complete model structure configuration, saved as a JSON format file.

[0181] Preferably, step S4 comprises the following steps:

[0182] Step S41: model input preprocessing is performed on the original data stream and the dynamic topology atlas to obtain a model input atlas;

[0183] Step S42: the model input atlas is input into the ST-GCN model to predict the future state, obtaining a predicted state sequence;

[0184] Step S43: monitoring point information extraction is performed on the dynamic topology graph to obtain monitoring point information; a preset safety threshold data and the monitoring point information are used to preliminarily judge the risk of the predicted state sequence to obtain an individual risk assessment result;

[0185] Step S44: based on the individual risk assessment result, a multi-parameter comprehensive risk assessment based on a Bayesian network is performed to obtain a comprehensive risk assessment result;

[0186] Step S45: a risk assessment report is generated for the predicted state sequence, the monitoring point information, the individual risk assessment result and the comprehensive risk assessment result to obtain a risk assessment report.

[0187] Step S46: a risk level is determined for the risk assessment report to obtain a determined risk level; a warning trigger condition is determined according to the determined risk level and a preset warning strategy configuration to obtain a warning trigger signal;

[0188] Step S47: warning information is generated according to the warning trigger signal, the risk assessment report and the monitoring point information to obtain to-be-sent warning information; the to-be-sent warning information is published to obtain a sent warning record.

[0189] In the embodiment of the application, the latest raw data stream (in CSV format) and dynamic topology graph data are read. The raw data stream is preprocessed: abnormal values are removed using the 3σ criterion, and missing values are filled using the linear interpolation method. Then, the data is scaled to the [0, 1] interval using the Min-Max normalization method. The preprocessed data is updated as node features to the dynamic topology graph. According to the input time step length T = 24 defined in step S3, data of 24 time steps before the current time t is extracted to construct a model input graph sequence. The sequence is converted to a tensor format as the input of the ST-GCN model. The tensor dimension is (1, T, N, F), where 1 represents a single sample, T is the time step length, N is the number of nodes, and F is the node feature dimension.

[0190] The ST-GCN model trained in step S36 is loaded. The model input graph sequence prepared in step S41 is input into the ST-GCN model for inference prediction. The model outputs a prediction result of P = 1 time steps in the future, i.e., a prediction value of 1 hour in the future. The prediction result is a tensor with a dimension of (1, P, N, F), where P is the prediction time step length, N is the number of nodes, and F is the node feature dimension. The tensor is converted to a time series format, containing the prediction value of each monitoring point within 1 hour in the future, and stored as a CSV file named the prediction state sequence.

[0191] Extract the monitoring point information from the dynamic topology map, including the monitoring point ID, location, monitoring parameter type, etc., and save it as a CSV file named Monitoring Point Information. Read the preset safety threshold data (CSV format), which contains the warning threshold and alarm threshold for each monitoring parameter. Traverse each monitoring point in the predicted state sequence and compare the predicted value with the corresponding warning threshold and alarm threshold. If the predicted value exceeds the warning threshold, mark the risk level of the monitoring point as "warning"; if the predicted value exceeds the alarm threshold, mark it as "alarm"; otherwise, mark it as "normal". Store the risk level and the magnitude of exceeding the threshold for each monitoring point into a CSV file named Individual Risk Assessment Results.

[0192] Load the pre-constructed Bayesian network model. This model defines the dependency relationships between different monitoring parameters and their impact on the overall risk level. Input the individual risk assessment results obtained in step S43 into the Bayesian network as evidence. Use the belief propagation algorithm for probabilistic reasoning to calculate the probability distribution of the overall risk level. Save the probability distribution of the risk level and the most likely risk level to a JSON file named Comprehensive Risk Assessment Results.

[0193] Read the predicted state sequence generated in step S42, the monitoring point information and individual risk assessment results generated in step S43, and the comprehensive risk assessment results generated in step S44. Use the pre-designed report template to generate a risk assessment report (PDF format). The report content includes: chart display of the predicted state sequence, risk level and magnitude of exceeding the threshold for each monitoring point, probability distribution and most likely risk level of the overall risk level, and specific risk description and recommended measures for high-risk monitoring points.

[0194] Read the risk assessment report generated in step S45, extract the most likely risk level in the comprehensive risk assessment results as the determined risk level. Read the preset warning strategy configuration (JSON format). This configuration defines the warning trigger conditions corresponding to different risk levels. For example, when the determined risk level is "alarm", immediately trigger the warning; when the determined risk level is "warning" and the duration exceeds 1 hour, trigger the warning. According to the determined risk level and the warning strategy configuration, judge whether the warning trigger condition is met. If it is met, generate a warning trigger signal containing the warning level (e.g. first-level warning, second-level warning) and the trigger reason.

[0195] If the pre-warning trigger signal is generated in step S46, a pre-warning information is generated according to the pre-warning trigger signal, the risk assessment report and the monitoring point information. The pre-warning information contains the pre-warning time, the pre-warning level, the trigger cause, the involved monitoring point position and monitoring parameter, the risk description and the suggested treatment measures. The pre-warning information is sent to a pre-defined receiver list, for example, by short message, email or special APP push. After the sending is completed, the pre-warning sending time, the receiver and the sending state (success or failure) are recorded and saved as a sent pre-warning record (CSV format).

[0196] Preferably, step S44 comprises the following steps:

[0197] Step S441: Obtain the water conservancy project safety evaluation index system and the historical risk event analysis report; define the basic monitoring parameter node based on the individual risk assessment result, to obtain the basic monitoring parameter node;

[0198] Step S442: Define the intermediate risk index node based on the basic monitoring parameter node and the water conservancy project safety evaluation index system, to obtain the intermediate risk index node;

[0199] Step S443: Define the final risk level node based on the intermediate risk index node, and construct the Bayesian network node definition table based on the basic monitoring parameter node, to obtain the Bayesian network node definition table;

[0200] Step S444: Establish the node directed edge based on the preset water conservancy professional knowledge base to obtain the node directed edge; model the causal relationship based on the Bayesian network node definition table to obtain the causal relationship data;

[0201] Step S445: Determine the network structure based on the Bayesian network node definition table using the historical risk event analysis report, to obtain the network structure data;

[0202] Step S446: Construct the Bayesian network structure based on the node directed edge, the causal relationship data and the network structure data, to obtain the Bayesian network structure diagram;

[0203] Step S447: Determine the conditional probability table based on the Bayesian network structure diagram, to obtain the Bayesian network conditional probability table;

[0204] Step S448: Input the individual risk assessment result as evidence into the Bayesian network structure diagram, and perform risk reasoning and assessment based on the Bayesian network conditional probability table, to obtain the comprehensive risk assessment result.

[0205] In the embodiment of the present application, a pre-defined water conservancy project safety evaluation index system document and a historical risk event analysis report are obtained. The safety evaluation index system document lists all relevant monitoring parameters, such as dam displacement, seepage pressure, reservoir water level, etc. The individual risk assessment result (in CSV format) generated in step S43 is read. Based on the individual risk assessment result and the safety evaluation index system, a basic monitoring parameter node is defined for each monitoring parameter. For example, if the individual risk assessment result contains the risk levels of dam crest displacement, dam body seepage pressure and reservoir water level, three basic monitoring parameter nodes of "dam crest displacement risk", "dam body seepage pressure risk" and "reservoir water level risk" are defined respectively. The names of all basic monitoring parameter nodes and the corresponding monitoring parameter types are recorded in a CSV file, named as basic monitoring parameter nodes.

[0206] The basic monitoring parameter nodes (in CSV format) generated in step S441 and the water conservancy project safety evaluation index system document are read. Based on the safety evaluation index system, relevant basic monitoring parameter nodes are combined into intermediate risk indicator nodes. For example, the two basic monitoring parameter nodes of "dam crest displacement risk" and "dam foundation settlement risk" are combined into an intermediate risk indicator node of "structural deformation risk"; the two basic monitoring parameter nodes of "dam body seepage pressure risk" and "drainage capacity risk" are combined into an intermediate risk indicator node of "seepage safety risk". The names of all intermediate risk indicator nodes, the corresponding basic monitoring parameter nodes, and the logical relationships (such as "and", "or") between them are recorded in a CSV file, named as intermediate risk indicator nodes.

[0207] The intermediate risk indicator nodes (in CSV format) generated in step S442 are read. A final risk level node is defined, such as "overall risk level". The values of this node can be pre-defined risk levels, such as "low", "medium", "high" or "normal", "attention", "warning", "alarm". The name of the final risk level node and the possible values are recorded in the CSV file. The information of the basic monitoring parameter nodes generated in step S441, the intermediate risk indicator nodes generated in step S442, and the final risk level node are integrated into a CSV file to form a Bayesian network node definition table. This table contains the names, types (basic monitoring parameters, intermediate risk indicators or final risk levels) and possible values of all nodes.

[0208] The Bayesian network node definition table generated in step S443 and the preset water conservancy professional knowledge base are read. The professional knowledge base is a database containing expert knowledge, for example, excessive displacement of the dam body will lead to increased structural risk, excessive seepage pressure will lead to increased seepage risk, etc. According to the professional knowledge base, directed edges are established between nodes in the Bayesian network node definition table. For example, a directed edge is established from the "dam crest displacement risk" node to the "structural deformation risk" node, indicating that the dam crest displacement risk will affect the structural deformation risk. The information of all established directed edges, including the starting node and the ending node, is saved to a CSV file named node directed edge. At the same time, the Bayesian network node definition table is modeled according to the cause-effect relationship in the field of water conservancy. For example, increased rainfall will lead to increased reservoir water level, and increased reservoir water level will lead to increased seepage pressure. These cause-effect relationships are expressed in mathematical formulas or logical expressions and saved to a text file named cause-effect relationship data.

[0209] The Bayesian network node definition table generated in step S443 and the historical risk event analysis report are read. The historical risk event analysis report records past risk events and the changes of various monitoring parameters before and after the risk events. According to the historical risk event analysis report, the relevance between various monitoring parameters and the degree of influence on risk events are analyzed. For example, if historical data shows that dam crest displacement and dam body seepage pressure both appear abnormal in multiple risk events, it is considered that there is strong relevance between the two parameters. These relevance information is used to determine the structure of the Bayesian network, for example, an edge is added between the "dam crest displacement risk" node and the "dam body seepage pressure risk" node, indicating the relevance between them. The final determined network structure information, including the connection relationship between nodes, is saved to a text file named network structure data.

[0210] The node directed edge (in CSV format), the cause-effect relationship data (in text format), and the network structure data (in text format) generated in step S445 are read. Using a Bayesian network modeling tool (such as GeNIe or BayesFusion), the Bayesian network structure is constructed based on these data. The node directed edge is imported into the modeling tool to create the connection relationship between nodes. According to the cause-effect relationship data, the network structure is adjusted to ensure correct expression of the cause-effect relationship. According to the network structure data, the network structure is further perfected, for example, adding or deleting the connection between nodes. The finally constructed Bayesian network structure is saved as a graph file (such as.xdsl format) named Bayesian network structure diagram.

[0211] The reading step S446 generates the Bayesian network structure diagram. For each node in the network, its conditional probability table (CPT) is determined. The CPT defines the probability distribution of the node when its parent nodes take different values. For the basic monitoring parameter nodes, the frequency of each risk level can be counted using historical monitoring data as its prior probability distribution. For the intermediate risk indicator nodes and the final risk level nodes, their CPTs can be determined in combination with expert knowledge and historical risk event data. The CPTs of all nodes are saved into a CSV file named Bayesian Network Conditional Probability Table. The probability values in the CPT should be represented using numerical values, such as 0.1, 0.5, 0.9, etc.

[0212] The reading step S446 generates the Bayesian network structure diagram, the step S447 generates the Bayesian network conditional probability table (in CSV format), and the step S43 generates the individual risk assessment result (in CSV format). The individual risk assessment result is input into the Bayesian network as evidence. For example, if the individual risk assessment result of the "dam crest displacement risk" node is "warning", the state of this node is set to "warning". The posterior probability distribution of the final risk level node is calculated using a Bayesian network inference algorithm (e.g., joint tree algorithm or variable elimination algorithm). The posterior probability distribution and the most likely risk level are saved into a JSON file named comprehensive risk assessment result.

[0213] Preferably, the present application also provides a data processing-based hydraulic engineering safety monitoring system for executing the data processing-based hydraulic engineering safety monitoring method as described above, which comprises:

[0214] A heterogeneous perception module is configured to construct a multi-source data acquisition scheme for the hydraulic engineering to obtain a multi-source data acquisition scheme, and to acquire multi-source data for the hydraulic engineering according to the multi-source data acquisition scheme to obtain an original data stream.

[0215] A topology coding module is configured to hierarchize monitoring nodes according to the multi-source data acquisition scheme to obtain an initial node set, to construct and define types of initial edges according to the initial node set to obtain an initial edge list, to assign weights to different types of initial edges in the initial edge list, and to generate an initial graph structure to obtain the initial graph structure, to analyze the correlation of the original data stream to obtain a monitoring point correlation matrix, and to adjust the dynamic edge weights of the initial graph structure using the monitoring point correlation matrix to obtain a dynamic topology graph.

[0216] A dynamic graph learning module is configured to obtain hydraulic engineering prediction demand data, to construct an ST-GCN model architecture according to the hydraulic engineering prediction demand data to obtain a model structure configuration, and to construct an ST-GCN model according to the model structure configuration and the dynamic topology graph, and to perform ST-GCN model training.

[0217] The intelligent early warning module is used for predicting future states by using the ST-GCN model to obtain a predicted state sequence, performing risk preliminary judgment on the predicted state sequence to obtain an individual risk assessment result, performing multi-parameter comprehensive risk assessment based on a Bayesian network according to the individual risk assessment result to obtain a comprehensive risk assessment result, performing risk assessment report generation according to the comprehensive risk assessment result to obtain a risk assessment report, performing early warning trigger condition judgment on the risk assessment report, and performing early warning information publishing to realize the water conservancy project safety monitoring task.

[0218] Therefore, from any viewpoint, the embodiments should be considered as being exemplary and non-limiting, the scope of the application being defined by the appended claims and not by the above description, and it is intended to embrace all variations falling within the meaning and the scope of the equivalent elements of the claims.

[0219] The above description is merely that of a specific implementation of the present application, enabling a person skilled in the art to understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments shown herein, but will accord with the widest scope consistent with the principles and novel features developed herein.

Claims

1. A data processing-based safety monitoring method for hydraulic engineering, characterized in that, The method comprises the following steps: Step S1: constructing a multi-source data collection scheme for the water conservancy project to obtain a multi-source data collection scheme; According to the multi-source data collection scheme, the water conservancy project is collected to obtain an original data stream; Step S2 comprises the following steps: Step S21: obtaining a water conservancy project structure design drawing; mapping the monitoring nodes of the multi-source data collection scheme and the water conservancy project structure design drawing to obtain a node mapping set; Step S22: classifying the node importance of the node mapping set to obtain an initial node set; Step S23 comprises the following steps: Step S231: marking the physical connection edges of the initial node set to obtain physical connection edges; marking the spatial proximity edges of the initial node set to obtain spatial proximity edges; Step S232: marking the water flow direction associated edges of the initial node set to obtain water flow associated edges; marking the stress transmission path associated edges of the initial node set to obtain stress associated edges; Step S233: constructing an initial edge list according to the physical connection edges, the spatial proximity edges, the water flow associated edges and the stress transmission path associated edges to obtain an initial edge list; Step S234: assigning weights to the physical connection edges based on material strength parameters to obtain physical connection edge weights; assigning weights to the spatial proximity edges based on the reciprocal of the Euclidean distance between nodes to obtain spatial proximity edge weights; Step S235: assigning weights to the water flow associated edges based on the fixed weight value of the directed edge to obtain water flow direction associated edge weights; assigning weights to the stress associated edges based on the stress transmission coefficient to obtain stress transmission path associated edge weights; Step S236: constructing a weighted initial edge list according to the physical connection edge weights, the spatial proximity edge weights, the water flow direction associated edge weights and the stress transmission path associated edge weights to obtain a weighted initial edge list; Step S24: generating an initial graph structure according to the weighted initial edge list and the initial edge list to obtain an initial graph structure; Step S25: performing correlation analysis on the original data stream based on historical data to obtain a monitoring point correlation matrix; performing original data preprocessing on the original data stream to obtain a preprocessed data stream; Step S26 comprises the following steps: Step S261: initializing the weights of the initial graph structure based on statistical correlation using the monitoring point correlation matrix to obtain a preliminary dynamic weight graph; Step S262: updating the real-time correlation based on a sliding window to obtain a real-time correlation matrix; Step S263: dynamically fusing and adjusting the preliminary dynamic weight graph according to the real-time correlation matrix to obtain an intermediate dynamic weight graph; Step S264: correcting the correlation based on the physical model to obtain a dynamic weighted graph; Step S27: performing data fusion on the dynamic weighted graph and the preprocessed data stream to obtain a dynamic topology graph; Step S3: Obtain water conservancy project prediction demand data; construct an ST-GCN model architecture based on the water conservancy project prediction demand data to obtain a model structure configuration; construct an ST-GCN model based on the model structure configuration and a dynamic topology graph, and optimize the ST-GCN model to obtain an ST-GCN model; Step S4: Use the ST-GCN model to predict future states to obtain a predicted state sequence; perform a risk preliminary judgment on the predicted state sequence to obtain an individual risk assessment result; perform a multi-parameter comprehensive risk assessment based on a Bayesian network based on the individual risk assessment result to obtain a comprehensive risk assessment result; generate a risk assessment report based on the comprehensive risk assessment result to obtain a risk assessment report; determine a warning trigger condition for the risk assessment report and publish warning information to achieve a water conservancy project safety monitoring task.

2. The waterwork safety monitoring method based on data processing according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Sensor deployment is performed on the water conservancy project, and a data transmission network is constructed to obtain a multi-source data acquisition scheme; Step S12: Multi-source data is collected from the water conservancy project according to the multi-source data acquisition scheme to obtain real-time multi-source data streams; Step S13: The real-time multi-source data streams are subjected to data format conversion and verification to obtain formatted and verified data; Step S14: The formatted and verified data are subjected to multi-source data integration and time synchronization to obtain an original data stream.

3. The waterwork safety monitoring method based on data processing according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Construct a model training data set based on a dynamic topology graph to obtain a model training data set; Step S32: Obtain water conservancy project prediction demand data; construct an ST-GCN model architecture based on the water conservancy project prediction demand data to obtain a model structure configuration; Step S33: Obtain prediction task type data; select a loss function and an optimizer based on the model structure configuration and the prediction task type data to obtain training parameter configuration; Step S34: Perform model training and verification based on the model structure configuration, the model training data set, and the training parameter configuration to obtain an initial ST-GCN model; Step S35: Perform model performance evaluation on the initial ST-GCN model to obtain a model evaluation report; Step S36: Optimize the model parameters of the initial ST-GCN model using the model evaluation report to obtain an ST-GCN model.

4. The water conservancy safety monitoring method based on data processing according to claim 3, characterized in that, Step S32 includes the following steps: Step S321: Generate a spatial convolution layer configuration based on Chebyshev spectrum convolution based on water conservancy project prediction demand data to obtain a spatial convolution layer configuration; Step S322: Construct a time convolution layer based on the spatial convolution layer configuration to obtain a time convolution layer configuration; Step S323: Integrate a time attention mechanism based on the spatial convolution layer configuration and the time convolution layer configuration to obtain an attention mechanism configuration; Step S324: Design a model layering structure based on the spatial convolution layer configuration, the time convolution layer configuration, and the attention mechanism configuration to obtain a model layering configuration; Step S325: Select an activation function for the model layering configuration to obtain an activation function configuration; Step S326: design the model output layer according to the model layer configuration and the activation function configuration to obtain a model structure configuration.

5. The data processing based safety monitoring method for hydraulic engineering according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: performing model input preprocessing on the original data stream and the dynamic topology atlas to obtain a model input atlas; Step S42: inputting the model input atlas into the ST-GCN model to perform future state prediction to obtain a predicted state sequence; Step S43: extracting monitoring point information from the dynamic topology atlas to obtain the monitoring point information; using a preset safety threshold data and the monitoring point information to perform risk preliminary judgment on the predicted state sequence to obtain an individual risk assessment result; Step S44: performing multi-parameter comprehensive risk assessment based on a Bayesian network according to the individual risk assessment result to obtain a comprehensive risk assessment result; Step S45: generating a risk assessment report on the predicted state sequence, the monitoring point information, the individual risk assessment result, and the comprehensive risk assessment result to obtain a risk assessment report; Step S46: determining a risk level for the risk assessment report to obtain a determined risk level; determining a warning trigger condition according to the determined risk level and a preset warning strategy configuration to obtain a warning trigger signal; Step S47: generating warning information according to the warning trigger signal, the risk assessment report, and the monitoring point information to obtain to-be-sent warning information; publishing the to-be-sent warning information to obtain a sent warning record.

6. The water conservancy project safety monitoring method based on data processing according to claim 5, characterized in that, Step S44 includes the following steps: Step S441: obtaining a water conservancy project safety evaluation index system and a historical risk event analysis report; defining a basic monitoring parameter node for the individual risk assessment result to obtain a basic monitoring parameter node; Step S442: defining an intermediate risk index node according to the basic monitoring parameter node and the water conservancy project safety evaluation index system to obtain an intermediate risk index node; Step S443: defining a final risk level node according to the intermediate risk index node, and constructing a Bayesian network node definition table according to the basic monitoring parameter node to obtain a Bayesian network node definition table; Step S444: establishing a node directed edge for the Bayesian network node definition table using a preset water conservancy professional knowledge base to obtain a node directed edge; modeling a causal relationship for the Bayesian network node definition table to obtain causal relationship data; Step S445: determining a network structure for the Bayesian network node definition table using a historical risk event analysis report to obtain network structure data; Step S446: constructing a Bayesian network structure according to the node directed edge, the causal relationship data, and the network structure data to obtain a Bayesian network structure diagram; Step S447: determining a conditional probability table for the Bayesian network structure diagram to obtain a Bayesian network conditional probability table; Step S448: inputting the individual risk assessment result as evidence into the Bayesian network structure diagram, and performing risk reasoning and assessment using the Bayesian network conditional probability table to obtain a comprehensive risk assessment result.

7. A data processing based safety monitoring system for hydraulic structures, characterized in that, The water conservancy project safety monitoring system is used for performing the water conservancy project safety monitoring method based on data processing. The isomerism perception module is used for constructing a multi-source data collection scheme for the water conservancy project to obtain a multi-source data collection scheme; and the water conservancy project is collected with multi-source data according to the multi-source data collection scheme to obtain an original data stream; The topology coding module is used for hierarchical monitoring node construction according to the multi-source data collection scheme to obtain an initial node set; initial edge construction and type definition are performed according to the initial node set to obtain an initial edge list; different types of initial edge weight assignment are performed on the initial edge list, and initial graph structure generation is performed to obtain the initial graph structure; correlation analysis is performed on the original data stream to obtain a monitoring point correlation matrix; and the initial graph structure is dynamically adjusted in weight by using the monitoring point correlation matrix to obtain a dynamic topology graph; The dynamic graph learning module is used for obtaining water conservancy project prediction demand data; ST-GCN model architecture construction is performed according to the water conservancy project prediction demand data to obtain a model structure configuration; an ST-GCN model is constructed according to the model structure configuration and the dynamic topology graph, and the ST-GCN model is optimized to obtain the ST-GCN model; The intelligent early warning module is used for predicting a future state by using the ST-GCN model to obtain a predicted state sequence; individual risk assessment results are obtained by performing risk preliminary judgment on the predicted state sequence; comprehensive risk assessment results are obtained by performing multi-parameter comprehensive risk assessment based on a Bayesian network according to the individual risk assessment results; a risk assessment report is generated according to the comprehensive risk assessment results to obtain a risk assessment report; early warning information is published by performing early warning trigger condition judgment on the risk assessment report, so as to realize the water conservancy project safety monitoring task.

Images