Hydropower station dam safety online monitoring method and system based on multi-source data

Abstract

The invention discloses a hydropower station dam safety online monitoring method and system based on multi-source data, and the method comprises the steps: obtaining the multi-source data information of a hydropower station dam, and carrying out the preprocessing and time sequence alignment of the multi-source data information, and obtaining dam monitoring standardized data; performing information transmission on the dam monitoring standardized data by adopting a convolutional neural network to obtain a weighted adjacency matrix, and extracting time sequence correlation characteristics of the dam monitoring standardized data; fusing the weighted adjacency matrix and the time sequence correlation features to obtain a dam feature matrix; key abnormal features are extracted based on the dam feature matrix; generating a monitoring result of the safety state of the hydropower station dam according to the key abnormal features; the key abnormal features are extracted from the hydropower station dam multi-source data, and the safety state of the hydropower station dam is accurately identified according to the key abnormal features.

Description

Technical Field

[0001] The present invention belongs to the technical field of dam monitoring, and in particular relates to a method and system for online monitoring of dam safety in a hydropower station. Background Art

[0002] Hydropower station dam safety monitoring is a key research area in the field of water conservancy engineering, directly impacting energy supply, the safety of people's lives and property, and the stability of the ecological environment. As the scale of hydropower development continues to expand, dam safety issues are becoming increasingly important, becoming a key component in ensuring the safety of national infrastructure.

[0003] Due to the diverse data sources for hydropower dams, including physical indicators such as structural stress, deformation, and seepage, as well as external environmental information such as hydrological and meteorological conditions, the heterogeneity and complexity of this data make unified modeling and analysis extremely difficult. However, current dam monitoring methods often rely on a single data source or static analysis, which makes it difficult to fully reflect the true state of the dam, resulting in the inability to effectively identify and address safety hazards. Summary of the Invention

[0004] The present invention provides a method and system for online monitoring of hydropower station dam safety based on multi-source data, which extracts key abnormal features from the multi-source data of the hydropower station dam and accurately identifies the safety status of the hydropower station dam according to the key abnormal features.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] A first aspect of the present invention provides a method for online monitoring of dam safety of a hydropower station based on multi-source data, comprising:

[0007] Obtain multi-source data information of the hydropower station dam, pre-process and time-series align the multi-source data information to obtain standardized dam monitoring data;

[0008] A convolutional neural network is used to transfer information on the standardized dam monitoring data to obtain a weighted adjacency matrix, and the temporal correlation features of the standardized dam monitoring data are extracted;

[0009] The weighted adjacency matrix and the time series correlation feature are fused to obtain a dam feature matrix; key abnormal features are extracted based on the dam feature matrix; and monitoring results of the safety status of the hydropower station dam are generated according to the key abnormal features.

[0010] Furthermore, the multi-source data information includes dam deformation monitoring data, dam stress state monitoring data, seepage monitoring data, hydrological and meteorological data, geological and topographic data, and maintenance and repair data;

[0011] The dam deformation monitoring data includes the dam's horizontal displacement data, vertical displacement data and deflection data; the dam's stress state monitoring data includes strain gauge data, vibrating wire stress gauge data and dam material stress and strain data; the seepage monitoring data includes seepage volume, seepage pressure and water quality; the hydrometeorological data includes water level, rainfall, air temperature and wind speed; the geological and topographic data includes the dam's geological structure, rock and soil properties and groundwater level.

[0012] Furthermore, the multi-source data information is pre-processed and time-series aligned to obtain standardized dam monitoring data, including:

[0013] A sliding window is constructed based on the fluctuation amplitude of multi-source data within a set time, and the sliding window is used to filter the multi-source data to eliminate high-frequency random noise in the multi-source data;

[0014] Linear interpolation is used to fill missing values ​​in multi-source data information. The expression formula is:

[0015]

[0016] In the formula, For multi-source data information Missing values ​​filled at all times; For multi-source data information Monitoring value at each moment; For multi-source data information Monitoring value at each moment; Missing values and monitoring values The step length between them; k is the monitoring value and monitoring values The number of missing values ​​between

[0017] Normalize the multi-source data information, and the expression formula is:

[0018]

[0019] In the formula, Multi-source data information The normalized features of For a single category The minimum value of For a single category The maximum value of

[0020] The normalized features of multi-source data information are time-series aligned and data grouped to obtain standardized data for dam monitoring.

[0021] Furthermore, data grouping is performed on the normalized features of the multi-source data information, and the specific process includes:

[0022] The K-Means clustering algorithm was used to group the dam deformation monitoring data, dam stress state monitoring data, seepage monitoring data, and maintenance and repair data according to monitoring locations to obtain location cluster labels. The location cluster labels were then added to the dam deformation monitoring data, dam stress state monitoring data, seepage monitoring data, and maintenance and repair data to obtain the dam spatial monitoring characteristics.

[0023] Based on hydrological and meteorological data as well as geological and topographic data, spatiotemporal feature vectors were constructed. DBSCAN density clustering was used to cluster the spatiotemporal feature vectors according to natural disaster events to obtain event cluster labels. Maintenance and inspection data were added to the hydrological and meteorological data as well as geological and topographic data to obtain dam event monitoring characteristics.

[0024] The dam monitoring standardized data is composed of dam spatial monitoring characteristics and dam event monitoring characteristics.

[0025] Furthermore, a convolutional neural network is used to transfer information from the standardized dam monitoring data to obtain a weighted adjacency matrix, which specifically includes:

[0026] The monitoring nodes are constructed based on the dam spatial monitoring characteristics in the standardized dam monitoring data, and the connection edges are constructed between each detection node based on the location cluster label to construct the monitoring node distribution map;

[0027] The convolutional neural network is set as a graph attention network. The feature graph is input into the pre-trained graph attention network to obtain the monitoring node feature vector. The Euclidean distance between the monitoring node feature vectors is calculated as the weight of the connection edge. The expression formula is:

[0028]

[0029] In the formula, is the edge connecting detection node i and detection node j; is the Euclidean distance between detection node i and detection node j, is the statistical standard deviation of the Euclidean distance; N is the number of monitoring nodes;

[0030] A weighted adjacency matrix is ​​constructed based on the weights of the monitoring nodes and the connecting edges.

[0031] Furthermore, extracting the temporal correlation features of the dam monitoring standardized data specifically includes:

[0032] Inputting the dam event monitoring features in the dam monitoring standardized data into the LSTM network, wherein the LSTM network includes an input gate, a forget gate, an update memory unit and an output gate;

[0033] The dam event monitoring feature is input into the forget gate to obtain the first memory superposition state, which is expressed as:

[0034]

[0035] In the formula, is the first memory superposition state at time t; Standardize dam event monitoring features in dam monitoring data; is the hidden memory state at time t-1; is the Sigmoid activation function; and is the weight matrix of the forget gate; is the bias vector of the forget gate;

[0036] The dam event monitoring feature is input into the input gate to obtain the first long-term memory state and the second memory superposition state, which is expressed as follows:

[0037]

[0038]

[0039] In the formula, is the second memory superposition state at time t, 、 、 and is the weight matrix of the input gate; and is the bias vector of the input gate; is the first long-term memory state at time t; is the hyperbolic tangent activation function;

[0040] The first long-term memory state, the first memory superposition state, and the second memory superposition state are input into the update memory unit to obtain the second long-term memory state, which is expressed as:

[0041]

[0042] In the formula, is the second long-term memory state at time t, is the second long-term memory state at time t-1;

[0043] The dam event monitoring feature and the hidden memory state are input into the output gate to obtain the hidden memory state, which is expressed as:

[0044]

[0045] In the formula, is the third memory superposition state at time t, and is the weight matrix of the output gate; is the bias vector of the output gate; is the hidden memory state at time t;

[0046] Repeat the iteration until all dam event monitoring features are input into the LSTM network, and use the hidden memory state output by the output gate in the LSTM network as the temporal association feature.

[0047] Furthermore, key abnormal features are extracted based on the dam feature matrix, specifically including:

[0048] The standard deviation and variance of each data point in the dam feature matrix are calculated, and the Laida criterion is used to set the anomaly threshold. The isolation forest algorithm is used to calculate the anomaly score of each data point in the dam feature matrix. When the anomaly score reaches the anomaly threshold, the key anomaly features of the dam feature matrix are output.

[0049] A second aspect of the present invention provides a hydropower station dam safety online monitoring system based on multi-source data, comprising:

[0050] The data acquisition module acquires multi-source data information of the hydropower station dam, pre-processes and time-series aligns the multi-source data information to obtain standardized dam monitoring data;

[0051] The association analysis module uses a convolutional neural network to perform information transmission on the dam monitoring standardized data to obtain a weighted adjacency matrix, and extracts the temporal correlation characteristics of the dam monitoring standardized data;

[0052] The anomaly judgment module fuses the weighted adjacency matrix and the time series correlation features to obtain a dam feature matrix; and extracts key anomaly features based on the dam feature matrix;

[0053] The output module generates monitoring results of the safety status of the hydropower station dam based on key abnormal characteristics.

[0054] The third aspect of the present invention provides an electronic device comprising a storage medium and a processor; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the online monitoring method for hydropower station dam safety described in the first aspect.

[0055] A fourth aspect of the present invention provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, and when the computer program is executed, the online monitoring method for hydropower station dam safety described in the first aspect is implemented.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] The present invention obtains multi-source data information of a hydropower station dam, preprocesses and time-aligns the multi-source data information to obtain standardized dam monitoring data; uses a convolutional neural network to transmit information on the standardized dam monitoring data to obtain a weighted adjacency matrix, which can fully consider the spatial relationships and mutual influences between data, extract the temporal correlation characteristics of the standardized dam monitoring data, and capture the changing trends and patterns of the data in the time dimension. This provides powerful intelligent auxiliary support for dam safety management and operation and maintenance decision-making.

[0058] The present invention fuses a weighted adjacency matrix with time-series correlation features to obtain a dam feature matrix; extracts key anomaly features based on the dam feature matrix; and generates monitoring results of the hydropower station dam safety status based on the key anomaly features. By mining the deep-level information and potential relationships hidden behind massive amounts of data, the dam feature matrix more accurately represents the actual state of the dam, thereby improving the reliability of the monitoring results. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 is a flow chart of the method for online monitoring of hydropower station dam safety provided in this embodiment 1; DETAILED DESCRIPTION

[0060] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention.

[0061] Example 1

[0062] like Figure 1 As shown, this implementation provides a hydropower station dam safety online monitoring method based on multi-source data, including

[0063] Acquiring multi-source data information of a hydropower station dam, the multi-source data information including dam deformation monitoring data, dam stress state monitoring data, seepage monitoring data, hydrological and meteorological data, geological and topographic data, and maintenance and repair data;

[0064] The dam deformation monitoring data includes the dam's horizontal displacement data, vertical displacement data and deflection data; the dam's stress state monitoring data includes strain gauge data, vibrating wire stress gauge data and dam material stress and strain data; the seepage monitoring data includes seepage volume, seepage pressure and water quality; the hydrometeorological data includes water level, rainfall, air temperature and wind speed; the geological and topographic data includes the dam's geological structure, rock and soil properties and groundwater level.

[0065] Preprocessing and time-series alignment of multi-source data information to obtain standardized dam monitoring data, including:

[0066] A sliding window is constructed based on the fluctuation amplitude of multi-source data within a set time, and the sliding window is used to filter the multi-source data to eliminate high-frequency random noise in the multi-source data;

[0067] Linear interpolation is used to fill missing values ​​in multi-source data information. The expression formula is:

[0068]

[0069] In the formula, For multi-source data information Missing values ​​filled at all times; For multi-source data information Monitoring value at each moment; For multi-source data information Monitoring value at each moment; Missing values and monitoring values The step length between them; k is the monitoring value and monitoring values The number of missing values ​​between

[0070] Normalize the multi-source data information, and the expression formula is:

[0071]

[0072] In the formula, Multi-source data information The normalized features of For a single category The minimum value of For a single category The maximum value of

[0073] First, the normalized features of multi-source data information are time-series aligned and data grouped. The specific process includes:

[0074] The K-Means clustering algorithm was used to group the dam deformation monitoring data, dam stress state monitoring data, seepage monitoring data, and maintenance and repair data according to monitoring locations to obtain location cluster labels. The location cluster labels were then added to the dam deformation monitoring data, dam stress state monitoring data, seepage monitoring data, and maintenance and repair data to obtain the dam spatial monitoring characteristics.

[0075] Based on hydrological and meteorological data as well as geological and topographic data, spatiotemporal feature vectors were constructed. DBSCAN density clustering was used to cluster the spatiotemporal feature vectors according to natural disaster events to obtain event cluster labels. Maintenance and inspection data were added to the hydrological and meteorological data as well as geological and topographic data to obtain dam event monitoring characteristics.

[0076] The dam monitoring standardized data is composed of dam spatial monitoring characteristics and dam event monitoring characteristics.

[0077] The K-Means clustering algorithm groups dam deformation, stress state, seepage, and maintenance and inspection data by monitoring location, assigning location cluster labels to form spatial monitoring features, accurately capturing the spatial correlations and characteristic differences between different dam parts. Furthermore, DBSCAN density clustering is used to classify spatiotemporal feature vectors constructed from hydrological, meteorological, geological, and topographic data by natural disaster events, generating event cluster labels. Maintenance and inspection data is then integrated to form event monitoring features. This allows for the extraction of deep-seated information and potential relationships hidden within massive amounts of data, such as the impact of natural disasters on dam safety and the impact of the spatial distribution of surface damage on dam safety.

[0078] The convolutional neural network is used to transfer information from the standardized dam monitoring data to obtain the weighted adjacency matrix, which includes:

[0079] The monitoring nodes are constructed based on the dam spatial monitoring characteristics in the standardized dam monitoring data, and the connection edges are constructed between each detection node based on the location cluster label to construct the monitoring node distribution map;

[0080] The convolutional neural network is set as a graph attention network. The feature graph is input into the pre-trained graph attention network to obtain the monitoring node feature vector. The Euclidean distance between the monitoring node feature vectors is calculated as the weight of the connection edge. The expression formula is:

[0081]

[0082] In the formula, is the edge connecting detection node i and detection node j; is the Euclidean distance between detection node i and detection node j, is the statistical standard deviation of the Euclidean distance; N is the number of monitoring nodes;

[0083] A weighted adjacency matrix is ​​constructed based on the weights of the monitoring nodes and the connecting edges.

[0084] The LSTM network is used to extract the temporal correlation features of the dam monitoring standardized data, including:

[0085] Inputting the dam event monitoring features in the dam monitoring standardized data into the LSTM network, wherein the LSTM network includes an input gate, a forget gate, an update memory unit and an output gate;

[0086] The dam event monitoring feature is input into the forget gate to obtain the first memory superposition state, which is expressed as:

[0087]

[0088] In the formula, is the first memory superposition state at time t; Standardize dam event monitoring features in dam monitoring data; is the hidden memory state at time t-1; is the Sigmoid activation function; and is the weight matrix of the forget gate; is the bias vector of the forget gate;

[0089] The dam event monitoring feature is input into the input gate to obtain the first long-term memory state and the second memory superposition state, which is expressed as follows:

[0090]

[0091]

[0092] In the formula, is the second memory superposition state at time t, 、 、 and is the weight matrix of the input gate; and is the bias vector of the input gate; is the first long-term memory state at time t; is the hyperbolic tangent activation function;

[0093] The first long-term memory state, the first memory superposition state, and the second memory superposition state are input into the update memory unit to obtain the second long-term memory state, which is expressed as:

[0094]

[0095] In the formula, is the second long-term memory state at time t, is the second long-term memory state at time t-1;

[0096] The dam event monitoring feature and the hidden memory state are input into the output gate to obtain the hidden memory state, which is expressed as:

[0097]

[0098] In the formula, is the third memory superposition state at time t, and is the weight matrix of the output gate; is the bias vector of the output gate; is the hidden memory state at time t;

[0099] Repeat the iteration until all dam event monitoring features are input into the LSTM network, and use the hidden memory state output by the output gate in the LSTM network as the temporal association feature.

[0100] The weighted adjacency matrix and the time series correlation features are fused to obtain the dam feature matrix. Based on the dam feature matrix, key abnormal features are extracted, including:

[0101] The standard deviation and variance of each data point in the dam characteristic matrix are calculated, and anomaly thresholds are set using the Laida criterion. Alternatively, anomaly thresholds are set based on industry standards, historical data, and expert knowledge to ensure they accurately reflect the range of characteristics of the dam under safe operating conditions. Weights are determined by comprehensively considering the importance and sensitivity of the characteristics. Using multi-criteria decision-making methods such as the Analytic Hierarchy Process, different characteristics are quantitatively weighted, highlighting the role of key characteristics in the comprehensive assessment of dam safety status.

[0102] The isolation forest algorithm is used to calculate the anomaly score of each data point in the dam feature matrix. When the anomaly score reaches the anomaly threshold, the key anomaly features of the dam feature matrix are output.

[0103] Because key abnormal characteristics of a dam do not exist in isolation, they often interact and influence each other in complex ways. For example, a dam's deformation characteristics may be closely coupled with its stress state characteristics and seepage characteristics. This embodiment identifies abnormal characteristics based on functional or causal relationships between these characteristics, predicting their potential impact on other related characteristics and thus enabling comprehensive, dynamic, and integrated monitoring of the dam's safety status.

[0104] Generate monitoring results of the hydropower station dam safety status based on key abnormal characteristics, including:

[0105] Based on the thresholds, weights, and feature association models set above, decision rules for dynamic control logic are constructed. When the key abnormal feature data monitored in real time exceeds the set threshold, the corresponding warning level is triggered according to the preset decision rules. Warning levels can be divided into multiple levels, such as minor warnings, moderate warnings, and severe warnings, and each warning level corresponds to different response measures. Minor warnings prompt operation and maintenance personnel to pay attention and further monitor relevant features, moderate warnings require on-site inspections and preliminary analysis, and severe warnings require the immediate activation of emergency plans and the implementation of emergency measures to ensure the safety of the dam. At the same time, the dynamic control logic can automatically adjust the warning status and response strategy based on the continuous updates and changes of real-time data, ensuring that the entire monitoring system is highly adaptable and real-time.

[0106] Example 2

[0107] This embodiment discloses an online monitoring system for the safety of a hydropower station dam based on multi-source data. The online monitoring system is used to execute the online monitoring method for the safety of a hydropower station dam described in Example 1. The online monitoring system includes:

[0108] The data acquisition module acquires multi-source data information of the hydropower station dam, pre-processes and time-series aligns the multi-source data information to obtain standardized dam monitoring data;

[0109] The association analysis module uses a convolutional neural network to transmit information on the dam monitoring standardized data to obtain a weighted adjacency matrix; and uses an LSTM network to extract the temporal correlation features of the dam monitoring standardized data;

[0110] The anomaly judgment module fuses the weighted adjacency matrix and the time series correlation features to obtain a dam feature matrix; and extracts key anomaly features based on the dam feature matrix;

[0111] The output module generates monitoring results of the safety status of the hydropower station dam based on key abnormal characteristics.

[0112] This embodiment obtains multi-source data information of a hydropower station dam, preprocesses and time-aligns the multi-source data information to obtain standardized dam monitoring data; uses a convolutional neural network to transmit information on the standardized dam monitoring data to obtain a weighted adjacency matrix, which can fully consider the spatial relationship and mutual influence between the data; uses an LSTM network to extract the temporal correlation features of the standardized dam monitoring data, which can capture the changing trends and patterns of the data in the time dimension; and provides powerful intelligent auxiliary support for dam safety management and operation and maintenance decision-making.

[0113] This embodiment fuses a weighted adjacency matrix and time series correlation features to obtain a dam feature matrix; extracts key abnormal features based on the dam feature matrix; generates monitoring results of the safety status of the hydropower station dam based on the key abnormal features; and by mining the deep information and potential relationships hidden behind the massive data, the dam feature matrix more accurately represents the actual status of the dam, thereby improving the reliability of the monitoring results.

[0114] Example 3

[0115] This embodiment provides an electronic device, including a storage medium and a processor; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the hydropower station dam safety online monitoring method described in Example 1.

[0116] Example 4

[0117] This embodiment provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, and when the computer program is executed, the hydropower station dam safety online monitoring method described in Example 1 is implemented.

[0118] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0119] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0120] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0121] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0122] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for online monitoring of hydropower station dam safety based on multi-source data, characterized in that: include: Obtain multi-source data information of the hydropower station dam, pre-process and time-series align the multi-source data information to obtain standardized dam monitoring data; A convolutional neural network is used to transfer information on the standardized dam monitoring data to obtain a weighted adjacency matrix, and the temporal correlation features of the standardized dam monitoring data are extracted; The weighted adjacency matrix and the time series correlation feature are fused to obtain a dam feature matrix; and key abnormal features are extracted based on the dam feature matrix; Generate monitoring results of the hydropower station dam safety status based on key abnormal characteristics.

2. The method for online monitoring of dam safety of a hydropower station according to claim 1, characterized in that: The multi-source data information includes dam deformation monitoring data, dam stress state monitoring data, seepage monitoring data, hydrological and meteorological data, geological and topographic data, and maintenance and repair data; The dam deformation monitoring data includes the dam's horizontal displacement data, vertical displacement data and deflection data; the dam's stress state monitoring data includes strain gauge data, vibrating wire stress gauge data and dam material stress and strain data; the seepage monitoring data includes seepage volume, seepage pressure and water quality; the hydrometeorological data includes water level, rainfall, air temperature and wind speed; the geological and topographic data includes the dam's geological structure, rock and soil properties and groundwater level.

3. The method for online monitoring of dam safety of a hydropower station according to claim 1, characterized in that: Preprocessing and time-series alignment of multi-source data information to obtain standardized dam monitoring data, including: A sliding window is constructed based on the fluctuation amplitude of multi-source data within a set time, and the sliding window is used to filter the multi-source data to eliminate high-frequency random noise in the multi-source data; Linear interpolation is used to fill missing values ​​in multi-source data information. The expression formula is: ； In the formula, For multi-source data information Missing values ​​filled at all times; For multi-source data information Monitoring value at each moment; For multi-source data information Monitoring value at each moment; Missing values and monitoring values The step length between them; k is the monitoring value and monitoring values The number of missing values ​​between Normalize the multi-source data information, and the expression formula is: ； In the formula, Multi-source data information The normalized features of For a single category The minimum value of For a single category The maximum value of The normalized features of multi-source data information are time-series aligned and data grouped to obtain standardized data for dam monitoring.

4. The method for online monitoring of dam safety of a hydropower station according to claim 1, characterized in that: Data grouping is performed based on the normalized features of multi-source data information. The specific process includes: The K-Means clustering algorithm was used to group the dam deformation monitoring data, dam stress state monitoring data, seepage monitoring data, and maintenance and repair data according to monitoring locations to obtain location cluster labels. The location cluster labels were then added to the dam deformation monitoring data, dam stress state monitoring data, seepage monitoring data, and maintenance and repair data to obtain the dam spatial monitoring characteristics. Based on hydrological and meteorological data as well as geological and topographic data, spatiotemporal feature vectors were constructed. DBSCAN density clustering was used to cluster the spatiotemporal feature vectors according to natural disaster events to obtain event cluster labels. Maintenance and inspection data were added to the hydrological and meteorological data as well as geological and topographic data to obtain dam event monitoring characteristics. The dam monitoring standardized data is composed of dam spatial monitoring characteristics and dam event monitoring characteristics.

5. The method for online monitoring of dam safety of a hydropower station according to claim 1, characterized in that: The convolutional neural network is used to transfer information from the standardized dam monitoring data to obtain the weighted adjacency matrix, which includes: The monitoring nodes are constructed based on the dam spatial monitoring characteristics in the standardized dam monitoring data, and the connection edges are constructed between each detection node based on the location cluster label to construct the monitoring node distribution map; The convolutional neural network is set as a graph attention network. The feature graph is input into the pre-trained graph attention network to obtain the monitoring node feature vector. The Euclidean distance between the monitoring node feature vectors is calculated as the weight of the connection edge. The expression formula is: ； In the formula, is the edge connecting detection node i and detection node j; is the Euclidean distance between detection node i and detection node j, is the statistical standard deviation of the Euclidean distance; N is the number of monitoring nodes; A weighted adjacency matrix is ​​constructed based on the weights of the monitoring nodes and the connecting edges.

6. The method for online monitoring of dam safety of a hydropower station according to claim 1, characterized in that: Extracting the temporal correlation features of the dam monitoring standardized data, specifically including: Inputting the dam event monitoring features in the dam monitoring standardized data into the LSTM network, wherein the LSTM network includes an input gate, a forget gate, an update memory unit and an output gate; The dam event monitoring feature is input into the forget gate to obtain the first memory superposition state, which is expressed as: ； In the formula, is the first memory superposition state at time t; Standardize dam event monitoring features in dam monitoring data; is the hidden memory state at time t-1; is the Sigmoid activation function; and is the weight matrix of the forget gate; is the bias vector of the forget gate; The dam event monitoring feature is input into the input gate to obtain the first long-term memory state and the second memory superposition state, which is expressed as follows: ； ； In the formula, is the second memory superposition state at time t, 、 、 and is the weight matrix of the input gate; and is the bias vector of the input gate; is the first long-term memory state at time t; is the hyperbolic tangent activation function; The first long-term memory state, the first memory superposition state, and the second memory superposition state are input into the update memory unit to obtain the second long-term memory state, which is expressed as: ； In the formula, is the second long-term memory state at time t, is the second long-term memory state at time t-1; The dam event monitoring feature and the hidden memory state are input into the output gate to obtain the hidden memory state, which is expressed as: ； In the formula, is the third memory superposition state at time t, and is the weight matrix of the output gate; is the bias vector of the output gate; is the hidden memory state at time t; Repeat the iteration until all dam event monitoring features are input into the LSTM network, and use the hidden memory state output by the output gate in the LSTM network as the temporal association feature.

7. The method for online monitoring of dam safety of a hydropower station according to claim 1, characterized in that: Key abnormal features are extracted based on the dam feature matrix, specifically including: The standard deviation and variance of each data point in the dam feature matrix are calculated, and the Laida criterion is used to set the anomaly threshold. The isolation forest algorithm is used to calculate the anomaly score of each data point in the dam feature matrix. When the anomaly score reaches the anomaly threshold, the key anomaly features of the dam feature matrix are output.

8. A hydropower station dam safety online monitoring system based on multi-source data, characterized by: include: The data acquisition module acquires multi-source data information of the hydropower station dam, pre-processes and time-series aligns the multi-source data information to obtain standardized dam monitoring data; The association analysis module uses a convolutional neural network to perform information transmission on the dam monitoring standardized data to obtain a weighted adjacency matrix, and extracts the temporal correlation characteristics of the dam monitoring standardized data; The anomaly judgment module fuses the weighted adjacency matrix and the time series correlation features to obtain a dam feature matrix; and extracts key anomaly features based on the dam feature matrix; The output module generates monitoring results of the safety status of the hydropower station dam based on key abnormal characteristics.

9. An electronic device comprising a storage medium and a processor; the storage medium is used to store instructions; characterized in that: The processor is configured to operate according to the instructions to execute the method for online monitoring of hydropower station dam safety according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed, the method for online monitoring of hydropower station dam safety according to any one of claims 1 to 7 is implemented.

Images