Water conservancy project risk point intelligent identification method and system based on internet of things

Abstract

The application relates to the technical field of water conservancy risk identification, and discloses a water conservancy engineering risk point intelligent identification method and system based on an Internet of Things. The method comprises the following steps: collecting sensing monitoring data, and transmitting the sensing monitoring data to an Internet of Things gateway; extracting data features related to water conservancy engineering risk factors from to-be-analyzed data, labeling risk point data, and constructing a risk identification model; performing time sequence modeling on all risk identification data, generating a risk prediction model, and outputting risk factor change results of any key monitoring area within a prediction time, and screening out first water conservancy risk points; quantitatively generating causal relationship strength in a causal relationship diagram based on the correlation of water conservancy engineering risk factors, screening out second water conservancy risk points in all key monitoring areas, and collecting the first water conservancy risk points and the second water conservancy risk points to generate a risk point list. The application improves the accuracy and comprehensiveness of water conservancy engineering risk point intelligent identification.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of water conservancy risk identification, in particular to a water conservancy engineering risk point intelligent identification method and system based on the Internet of Things. BACKGROUND

[0002] Water conservancy engineering covers multiple fields such as flood control, irrigation, water supply, and power generation. However, water conservancy engineering faces many risks in the operation process, such as dam damage caused by flood erosion, structural aging of water conservancy structures, leakage, damage caused by natural disasters such as earthquakes, and equipment failure caused by human factors. Traditional water conservancy engineering risk identification mainly relies on manual inspection. Although some water conservancy projects have installed some monitoring equipment such as water level meters, rain gauges, and osmotic pressure meters, these devices are often arranged in a scattered manner and lack effective integration and unified management. The data of each monitoring device is independently stored and transmitted, making it difficult to form a complete risk assessment system. Therefore, it is necessary to combine Internet of Things technology and artificial intelligence technology to provide new ideas and technical means for water conservancy engineering risk identification, effectively overcome the limitations of traditional methods, and realize real-time, accurate, and intelligent identification of water conservancy engineering risks.

[0003] Similar prior art includes Chinese patent application No. CN119809327A, which discloses a water conservancy channel engineering risk point automatic identification method and system, relating to the technical field of engineering risk identification, including an engineering information acquisition module, a construction personnel analysis module, an engineering risk point preliminary analysis module, an environment detection module, an engineering risk point comprehensive analysis module, and a database. The basic information and work information of the construction personnel during the target water conservancy channel engineering process are acquired and comprehensively analyzed to obtain the engineering risk coefficient corresponding to the construction position of each construction personnel. The environmental factors and hydrological factors of the target water conservancy channel engineering are comprehensively analyzed to obtain the confirmed engineering risk points corresponding to the target water conservancy channel engineering. The engineering risk points of the target water conservancy channel engineering are effectively and quickly locked according to the environmental conditions and construction conditions. Chinese patent application No. CN119887028A discloses a water conservancy engineering risk assessment method and system based on artificial intelligence, relating to the technical field of risk assessment management. The method includes: acquiring a feature parameter set; invoking an information value quantization function to analyze the prediction ability quantization of the water conservancy engineering risk coefficient to obtain a first value coefficient; if the first value coefficient reaches a predetermined threshold, adding the first feature to a candidate feature factor list; activating an accident causation Bayesian network model to perform causation analysis on the candidate feature factor list to obtain a target causation chain and a target key feature factor list; collecting dynamic features of the target water conservancy engineering to obtain a real-time feature parameter set; performing normalization calculation and analysis to obtain the target risk coefficient of the target water conservancy engineering.

[0004] The deficiencies of the prior art mainly manifest in that the data collection dimension is single, the multi-source heterogeneous data is not fully integrated, the identification result of the engineering risk point is single, and in actual conditions, the water conservancy engineering risk point and the type of risk factors need to be accurately identified through the fusion monitoring of various sensing devices. SUMMARY

[0005] The application provides a water conservancy engineering risk point intelligent identification method and system based on Internet of Things, which is used for improving the accuracy and comprehensiveness of the water conservancy engineering risk point intelligent identification based on Internet of Things.

[0006] In a first aspect, the application provides a water conservancy engineering risk point intelligent identification method based on Internet of Things, which comprises:

[0007] An area distribution map of a water conservancy project is acquired, and multiple sensors are deployed in key monitoring areas of the water conservancy project based on the area distribution map. The sensors collect sensing monitoring data based on a preset collection mode, and transmit the sensing monitoring data to an Internet of Things gateway. The Internet of Things gateway pre-processes the received sensing monitoring data, and stores the sensing monitoring data classified based on the key monitoring areas and sensor types to a cloud server.

[0008] The sensing monitoring data contained in any key monitoring area is set as to-be-analyzed data, data features related to water conservancy engineering risk factors are extracted from the to-be-analyzed data, risk point data existing is labeled based on the data features, a risk identification model is constructed based on the labeled data features, and the risk identification model is deployed to the cloud server.

[0009] The cloud server receives real-time monitoring data of any key monitoring area, the risk identification model outputs risk identification data based on the real-time monitoring data, all the risk identification data are time series modeled to generate a risk prediction model, the risk prediction model outputs risk factor change results of any key monitoring area in a prediction time, and a first water conservancy risk point is screened out from all the key monitoring areas based on the risk factor change results.

[0010] A causal relationship graph of all the key monitoring areas is constructed based on the area distribution map, a causal relationship strength is quantitatively generated in the causal relationship graph based on the correlation of the water conservancy engineering risk factors, a second water conservancy risk point is screened out from all the key monitoring areas based on the causal relationship strength, and a risk point list is generated by aggregating the first water conservancy risk point and the second water conservancy risk point based on the water conservancy engineering risk factors.

[0011] In combination with the first aspect, the transmitting the sensing monitoring data to the Internet of Things gateway comprises:

[0012] connecting all the sensors and the Internet of Things gateway to the same Internet of Things network, configuring communication parameters for the sensors, and setting unique node IDs for all the sensors and the Internet of Things gateway;

[0013] In collecting the sensing monitoring data, the sensor checks whether there is valid routing information reaching the Internet of Things gateway, if there is, the sensing monitoring data is transmitted to the Internet of Things gateway based on the reaching path of the valid routing information, if not, an optimal path is generated in the Internet of Things network, and the sensing monitoring data is transmitted to the Internet of Things gateway based on the optimal path.

[0014] In combination with the first aspect, the generating an optimal path in the Internet of Things network comprises:

[0015] The sensor transmits routing request information to all adjacent nodes;

[0016] After the adjacent node receives the routing request information, it checks whether the routing request information has passed through the current node, if the current node is the destination node corresponding to the Internet of Things gateway, it enters the response processing stage, otherwise, the ID tag of the current node is added to the request tag corresponding to the routing request information, and the routing request information is transmitted to the remaining adjacent nodes;

[0017] After the destination node receives the routing request information, it generates routing response information and adds its node ID to the routing response information, and the routing response information returns to the sensor along the reverse direction of the routing request information;

[0018] After the intermediate node receives the routing response information, it extracts the tag information, generates an extended tag, adds the extended tag to the routing response information, and continues to send the routing response information to the sensor, the intermediate node takes the received tag information as a basic tag and generates a new extended tag for the next node;

[0019] After the sensor receives multiple routing response information, the path represented by the longest tag information is set as the optimal path based on the length of the tag.

[0020] In combination with the first aspect, the extracting data features related to water conservancy risk factors from the to-be-analyzed data comprises:

[0021] Extracting risk factor data from the data to be analyzed based on the water conservancy project risk factors, wherein the water conservancy project risk factors include water level factors, leakage factors, structural deformation factors, and flow rate factors;

[0022] Establishing a physical model based on the sensor type, normalizing the risk factor data and outputting it to the physical model, calculating a ratio of a difference between adjacent time points of the risk factor data and a time interval, setting the ratio as a risk change rate, calculating a difference between a maximum value and a minimum value of the risk factor data within a preset time window, summarizing all the differences and setting them as a fluctuation value distribution, and setting the risk change rate and the fluctuation value distribution as physical features;

[0023] Establishing a sequence analysis model based on the data type, wherein the sequence analysis model extracts time series features from the risk factor data, wherein the time series features include autocorrelation coefficient, partial autocorrelation coefficient and wavelet transform features;

[0024] Establishing an image analysis model, inputting the risk factor data into the image analysis model based on the data type, and extracting image features, wherein the image features include texture features, edge detection features, and target monitoring features;

[0025] The physical features, the time series features, and the image features are summarized based on the categories of the water conservancy project risk factors and set as the data features.

[0026] In combination with the first aspect, the step of constructing a risk identification model based on the labeled data features includes:

[0027] If the data feature corresponding to any of the water conservancy project risk factors does not meet the preset threshold, the data to be analyzed corresponding to the data feature is set as the risk point data;

[0028] Constructing a machine learning model, inputting the data features corresponding to all the key monitoring areas into the machine learning model for classification training, wherein the machine learning model outputs an identification result based on the risk point data, wherein the identification result includes a risk point category, a risk point indicator, and a risk point confidence level;

[0029] The parameters of the machine learning model are adjusted based on the confidence of the risk point, and the adjusted machine learning model is set as the risk identification model.

[0030] In combination with the first aspect, the risk prediction model outputs the risk factor change results of any key monitoring area within the prediction time, including:

[0031] Based on data correlation, the risk identification data corresponding to any of the key monitoring areas are sequentially combined to generate multiple risk sequence groups, and the risk sequence groups contained in all the key monitoring areas are pre-processed and classified into training sets and test sets;

[0032] Using a deep learning framework to construct the risk prediction model, inputting the training set into the risk prediction model for training, using the test set to evaluate the trained risk prediction model, and calculating the prediction error;

[0033] The risk category label of the risk sequence group is set based on the sensor type, the risk sequence group contained in any of the key monitoring areas is input into the risk prediction model, the change sequence within the prediction time is output based on the risk category label, the change sequence is corrected based on the prediction error and set as the risk factor change result.

[0034] In combination with the first aspect, if there are discrete points in the risk factor change results that are greater than the first preset value, any of the key monitoring areas will be set as the first water conservancy risk point, and the risk category label corresponding to the discrete point will be set as the risk factor.

[0035] In combination with the first aspect, screening out the second water conservancy risk point in all the key monitoring areas based on the strength of the causal relationship includes:

[0036] Setting any of the key monitoring areas as a regional node, setting directed edges based on the operation layout between the nodes, and connecting all the regional nodes using the directed edges to generate the causal relationship graph;

[0037] Splitting the causal relationship graph into multiple sub-graphs based on a directed connection method, if the regional node corresponding to the first water conservancy risk point exists in the sub-graph, establishing a Bayesian network structure based on the sub-graph, and estimating the conditional probability distribution of the Bayesian network structure using the real-time monitoring data corresponding to the regional node in the sub-graph to generate the causal relationship strength;

[0038] If the causal relationship strength is greater than a second preset value, the regional sub-node of the first water risk point in the sub-relationship graph is set as the second water risk point.

[0039] In a second aspect, the present application provides an intelligent identification system for water conservancy project risk points based on the Internet of Things, the intelligent identification system for water conservancy project risk points based on the Internet of Things comprising:

[0040] The data transmission module is configured to acquire a regional distribution map of the water conservancy project, deploy a plurality of sensors in key monitoring areas of the water conservancy project based on the regional distribution map, collect sensing monitoring data based on a preset collection mode, and transmit the sensing monitoring data to an Internet of Things gateway.

[0041] The risk identification module is configured to set the sensing monitoring data contained in any key monitoring area as to-be-analyzed data, extract data features related to water conservancy risk factors from the to-be-analyzed data, label risk point data existing based on the data features, construct a risk identification model based on the labeled data features, and deploy the risk identification model to the cloud server.

[0042] The risk prediction module is configured to receive real-time monitoring data of any key monitoring area by the cloud server, output risk identification data based on the real-time monitoring data by the risk identification model, perform time series modeling on all the risk identification data, generate a risk prediction model, output risk factor change results of any key monitoring area within a prediction time by the risk prediction model, and screen a first water conservancy risk point from all the key monitoring areas based on the risk factor change results.

[0043] The summary module is configured to construct a causal relationship diagram of all the key monitoring areas according to the regional distribution map, quantitatively generate a causal relationship strength in the causal relationship diagram based on the correlation of the water conservancy risk factors, screen a second water conservancy risk point from all the key monitoring areas based on the causal relationship strength, summarize the first water conservancy risk point and the second water conservancy risk point based on the water conservancy risk factors, and generate a risk point list.

[0044] In the technical solution provided by this application, first, a variety of sensors are deployed in the key monitoring areas of water conservancy projects using Internet of Things technology. The collected monitoring data is pre-processed through the Internet of Things gateway and stored in a cloud server in a classified manner. Through the interaction of routing request information and routing response information, the sensor can quickly find the optimal path to the Internet of Things gateway. The collaboration of adjacent nodes and the dynamic update of label information further optimize the path selection process, ensuring the stability of the network and the accuracy of data transmission. Then, data features are extracted from three dimensions: physical model, time series model and image model, which comprehensively reflects the dynamic changes, time correlation and spatial distribution of water conservancy project risk factors. By calculating the risk change rate and fluctuation value distribution, the change rate and fluctuation of risk factors are quantified, providing an important physical basis for risk identification. The autocorrelation coefficient, partial autocorrelation coefficient and wavelet transform features are extracted to reflect the long-term trend and periodic changes of risk factors, which helps to predict the development trend of risks. The texture features, edge detection features and target monitoring features are extracted to intuitively reflect the spatial distribution and morphological changes of risk factors. Different types of features are summarized according to the categories of water conservancy project risk factors to form a complete data feature set, which provides rich feature input for the subsequent risk identification model construction. Finally, a suitable machine learning model is selected to classify and identify risk points in water conservancy projects. It can learn the complex relationship between data features and risk points, realize automatic identification and classification of risk points, and use a deep learning framework to build a risk prediction model. It can learn the complex patterns and dynamic changes of risk sequence groups, improve the accuracy of risk prediction, and realize intelligent prediction of risk factors through deep learning models, reduce manual intervention, and improve the automation level of water conservancy project risk monitoring.

[0045] This application also constructs a causal relationship diagram to intuitively represent the mutual influence relationship between key monitoring areas, establishes a Bayesian network structure based on the sub-relationship diagram, and estimates the conditional probability distribution, which can quantify the causal relationship between variables, improve the scientificity and accuracy of risk point screening, calculate the causal relationship strength through the Bayesian network, quantify the degree of mutual influence between nodes in each area, and provide a clear quantitative standard for screening the second water conservancy risk point. The dynamic calculation of the causal relationship strength through the Bayesian network and real-time monitoring data can adapt to the dynamic changes in the operation of water conservancy projects and improve the intelligence level of risk monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0047] Figure 1 This is a schematic diagram of an embodiment of a method for intelligently identifying risk points in water conservancy projects based on the Internet of Things in an embodiment of the present application;

[0048] Figure 2 This is a schematic diagram of an embodiment of the process of extracting features from risk factor data in an embodiment of the present application;

[0049] Figure 3 This is a schematic diagram of an embodiment of the risk identification model construction and evaluation process in the embodiment of the present application;

[0050] Figure 4 This is a schematic diagram of an embodiment of the intelligent identification system for risk points of water conservancy projects based on the Internet of Things in the embodiment of this application. DETAILED DESCRIPTION

[0051] The embodiments of the present application provide a method and system for intelligent identification of risk points in water conservancy projects based on the Internet of Things. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or that are inherent to these processes, methods, products or devices.

[0052] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiment of the present application, an embodiment of the method for intelligently identifying risk points of water conservancy projects based on the Internet of Things includes:

[0053] Step S101: Obtain a regional distribution map of the water conservancy project, and deploy multiple sensors in the key monitoring areas of the water conservancy project based on the regional distribution map. The sensors collect sensor monitoring data based on a preset collection method and transmit the sensor monitoring data to the Internet of Things gateway. The Internet of Things gateway pre-processes the received sensor monitoring data and classifies the sensor monitoring data based on the key monitoring area and sensor type and stores it in the cloud server.

[0054] It is understandable that the execution subject of this application can be an intelligent identification device for water conservancy project risk points based on the Internet of Things, or a terminal or a server, which is not limited here. The embodiment of this application is described by taking the server as the execution subject as an example.

[0055] Specifically, the regional distribution map includes key areas of the water conservancy project (such as dams, spillways, sluice gates, and diversion channels). Based on this regional distribution map, key monitoring areas are identified. Key monitoring areas are typically risk-prone areas of the water conservancy project, such as the dam body, foundation, and abutments, as well as the entrances and exits of the spillway. Sensors include, but are not limited to, water level sensors, seepage sensors, stress and strain sensors, and meteorological sensors. Sensors collect monitoring data according to preset collection methods (such as timed collection or event-triggered collection) and transmit the data to the IoT gateway via wireless or wired communication. Preprocessing includes operations such as data cleaning (removing noise and abnormal data) and data formatting (unifying data formats). Preprocessed monitoring data is categorized and stored in a cloud server based on key monitoring areas and sensor types. For example, within the same key monitoring area, water level sensor data is stored in the "water level monitoring data" category, while seepage sensor data is stored in the "seepage monitoring data" category. This categorized storage facilitates subsequent data analysis and processing.

[0056] Step S102: Set the sensor monitoring data contained in any key monitoring area as the data to be analyzed, extract data features related to the risk factors of water conservancy projects from the data to be analyzed, mark the existing risk point data based on the data features, build a risk identification model based on the marked data features, and deploy the risk identification model to the cloud server.

[0057] Specifically, risk factors for water conservancy projects include but are not limited to water level factors, leakage factors, structural deformation factors, and flow rate factors. Data features are extracted for the data to be analyzed according to different data categories, and the specific extraction methods are described later. For example, for dam seepage monitoring data, features such as seepage rate and seepage pressure can be extracted; for water level monitoring data, features such as water level change rate and water level peak can be extracted. Based on the extracted data features, the existing risk point data are annotated. For example, if the seepage rate in a key monitoring area exceeds the set threshold, it is marked as a potential seepage risk point. Then, a risk identification model is constructed based on the annotated data features and deployed to the cloud server so that it can receive real-time monitoring data of key monitoring areas in real time and output risk identification data, where risk identification data refers to data that contains risks in the real-time monitoring data, as well as the risk type.

[0058] Step S103: The cloud server receives the real-time monitoring data of any key monitoring area. The risk identification model outputs the risk identification data based on the real-time monitoring data. All the risk identification data are modeled as time series to generate a risk prediction model. The risk prediction model outputs the risk factor change results of any key monitoring area within the prediction time. Based on the risk factor change results, the first water conservancy risk point is screened out in all key monitoring areas.

[0059] Specifically, the time series modeling can adopt the LSTM neural network method to predict the risk factor changes of the key monitoring areas in the future period of time. The risk prediction model outputs the risk factor change results of the key monitoring areas in the prediction time, and according to these results, the first water conservancy risk point is screened out among all the key monitoring areas. For example, if the prediction result shows that the seepage rate of a certain area will continue to increase and exceed the safety threshold in the future period of time, it will be screened as the first water conservancy risk point, and “excessive seepage” as the risk factor.

[0060] Step S104, constructing a causal relationship graph of all key monitoring areas based on the area distribution map, quantifying the causal relationship strength in the causal relationship graph based on the correlation of water conservancy engineering risk factors, screening the second water conservancy risk point among all key monitoring areas based on the causal relationship strength, and summarizing the first water conservancy risk point and the second water conservancy risk point based on the water conservancy engineering risk factors to generate a risk point list.

[0061] Specifically, the causal relationship graph reflects the mutual influence relationship between each key monitoring area, for example, the increase of rainfall may cause the water level of the reservoir to rise, and then affect the seepage of the dam. The causal relationship strength refers to the probability value between the causal relationships, for example, if the rainfall of a certain area increases, and the causal relationship strength between it and the downstream dam is high, the downstream dam will be screened as the second water conservancy risk point. If a certain area is screened as the first water conservancy risk point (based on real-time monitoring data prediction) and the second water conservancy risk point (based on causal relationship analysis) at the same time, it will be marked as a high-risk area in the summary. The generation of the risk point list includes the location, risk factor, risk level and other information of all screened risk points, providing decision support for the management personnel of the water conservancy project.

[0062] In a specific embodiment, the sensing monitoring data is transmitted to the Internet of Things gateway, comprising:

[0063] (1) connecting all sensors and the Internet of Things gateway to the same Internet of Things network, configuring communication parameters for the sensors, and setting unique node IDs for all sensors and the Internet of Things gateway.

[0064] (2) When collecting and generating sensing monitoring data, the sensor checks whether there is valid routing information to the Internet of Things gateway, if there is, the sensing monitoring data is transmitted to the Internet of Things gateway based on the arrival path of the valid routing information, if not, an optimal path is generated in the Internet of Things network, and the sensing monitoring data is transmitted to the Internet of Things gateway based on the optimal path.

[0065] Specifically, the Internet of Things network can be a network based on ZigBee, LoRa, NB-IoT, etc. Low-power wide-area network (LPWAN) technology, or a network based on Wi-Fi or 4G / 5G, the specific choice depends on the geographical location and monitoring requirements of the water conservancy project. Configure communication parameters for each sensor, including communication frequency, data transmission rate, encryption method, etc. These communication parameters need to be optimized according to the type of sensor and network environment to ensure the stability and security of data transmission. Assign a unique node ID to each sensor and Internet of Things gateway, which is used to uniquely identify each device in the network, facilitating data transmission and routing management.

[0066] When the sensor collects the sensing monitoring data, first check if there is valid routing information to the Internet of Things gateway, valid routing information refers to the path information that has been established and can ensure successful data transmission to the Internet of Things gateway. If there is no valid routing information, the sensor will generate an optimal path in the Internet of Things network.

[0067] In a specific embodiment, generating an optimal path in the Internet of Things network includes:

[0068] (1) The sensor transmits routing request information to all adjacent nodes.

[0069] (2) After the adjacent nodes receive the routing request information, check if the routing request information has passed through the current node, if the current node is the destination node corresponding to the Internet of Things gateway, enter the response processing stage, otherwise, add the ID tag of the current node to the request tag corresponding to the routing request information, and transmit the routing request information to the remaining adjacent nodes.

[0070] (3) After the destination node receives the routing request information, generate routing response information and add its node ID to the routing response information, and the routing response information returns to the sensor along the reverse direction of the routing request information.

[0071] (4) After the intermediate node receives the routing response information, extract the tag information, generate an extended tag, add the extended tag to the routing response information, and continue to send the routing response information to the sensor, the intermediate node takes the received tag information as the basic tag, and generates a new extended tag for the next node.

[0072] (5) After the sensor receives multiple routing response information, set the path represented by the longest tag information as the optimal path based on the length of the tag.

[0073] Specifically, the distribution distance between nodes is determined according to the distribution coordinates corresponding to the node IDs, and the adjacent nodes can be found according to the distribution distance. The routing request information contains the node ID of the sensor (source node), the node ID of the target node (Internet of Things gateway), the request label (an empty label list used to record the node ID passed through), the hop count (initial value is 0, used to record the hop count from the source node to the current node), and other information. The routing request information is used to find the path to the Internet of Things gateway, and the routing response information is used to return the path information. After receiving the routing request information, the adjacent node first checks whether the request label in the routing request information contains its own node ID. The response processing stage refers to the data transmission stage. By checking whether the request label contains the ID of the current node, the routing request information is prevented from circulating between nodes.

[0074] After receiving the routing request information, the Internet of Things gateway (destination node) generates a routing response information, including the node ID of the Internet of Things gateway (source node), the node ID of the sensor (destination node), and the response label (containing the node ID of the Internet of Things gateway). The return path of the routing response information is the same as the path of the routing request information, but the direction is opposite.

[0075] The label information is extracted from the request label, and the intermediate node generates an extended label according to the label information in the routing response information and adds the extended label to the routing response information. By dynamically updating the label information, it is ensured that the routing response information can return to the sensor along the correct path.

[0076] The sensor can receive multiple routing response information, because there can be multiple paths to reach the Internet of Things gateway. The path represented by the routing response information with the longest label length is selected as the optimal path. The longer the label length, the more nodes the path passes through, and the more complete the path. The above steps optimize the transmission process of the sensor monitoring data, improve the reliability and efficiency of the Internet of Things network in water conservancy engineering monitoring, and provide more efficient data transmission support for intelligent identification of water conservancy risk points.

[0077] In a specific embodiment, data features related to water conservancy risk factors are extracted from the data to be analyzed, including:

[0078] (1) Risk factor data is extracted from the data to be analyzed based on water conservancy risk factors, wherein the water conservancy risk factors include water level factors, leakage factors, structural deformation factors, and flow rate factors.

[0079] (2) Establish a physical model based on the sensor type, standardize the risk factor data and output it into the physical model, calculate the ratio of the difference value of the risk factor data at adjacent time points to the time interval, set the ratio as the risk change rate, calculate the difference value between the maximum value and the minimum value of the risk factor data within the preset time window, aggregate all the difference values and set them as the fluctuation value distribution, and set the risk change rate and the fluctuation value distribution as the physical characteristics.

[0080] (3) Establish a sequence analysis model based on the data type, and the sequence analysis model extracts time series features of the risk factor data, including autocorrelation coefficients, partial autocorrelation coefficients and wavelet transform features.

[0081] (4) Establish an image analysis model, input the risk factor data into the image analysis model based on the data type, extract image features, and the image features include texture features, edge detection features and target monitoring features.

[0082] (5) Based on the category of the risk factors of the water conservancy project, the physical characteristics, time series features and image features are aggregated and set as data features.

[0083] Specifically, Figure 2 The flowchart for feature extraction of risk factor data. The water level factor refers to the information of reservoir water level, river water level, etc., the seepage factor refers to the information of dam seepage rate, seepage pressure, etc., the structural deformation factor refers to the information of dam displacement, settlement, etc., and the flow velocity factor refers to the information of water flow velocity, flood discharge flow velocity, etc. According to the sensor type and the key monitoring area, the data related to the above risk factors, i.e. the risk factor data, are extracted from the data to be analyzed, for example, the water level data is extracted from the water level sensor, and the seepage data is extracted from the seepage sensor.

[0084] The sensor type refers to the category of the sensor. According to the sensor type, a physical model is established, for example, for a water level sensor, a physical model of water level change is established; for a seepage sensor, a physical model of seepage change is established. The data standardization processing is to standardize the risk factor data, eliminate the influence of different dimensions and orders of magnitude, for example, all data are normalized to the interval [0, 1]. The calculation formula of the risk change rate P1 is: , wherein, is the risk factor data at the current time point, is the risk factor data at the previous adjacent time point, is the time interval. Setting the risk change rate and the fluctuation value distribution as the physical characteristics can clearly obtain the physical fluctuation change contained in the risk factor data.

[0085] Sequence analysis models refer to time series analysis models. For example, the ARIMA model can be used for water level data, while the wavelet transform model can be used for seepage data. The autocorrelation coefficient reflects the correlation of data at different time lags. The partial autocorrelation coefficient reflects the correlation of data after removing the influence of intermediate lags. By performing a wavelet transform on risk factor data and extracting the wavelet coefficients, we can reflect the characteristics of the data at different frequencies. Therefore, sequence analysis models can extract time series features from risk factor data that contain temporal characteristics.

[0086] Risk factor data may contain images. We use a convolutional neural network (CNN) to build an image analysis model and extract image features from risk factor data. We can use methods like the gray-level co-occurrence matrix to extract texture features, edge detection algorithms like Canny to extract edge features, and object detection algorithms (such as YOLO) to extract target features in images. For example, to monitor dam cracks, these features can be used as image features.

[0087] Based on the categories of water conservancy project risk factors (such as water level factors and leakage factors), physical characteristics, time series characteristics, and image characteristics are aggregated to form a complete data feature set. For example, for water level factors, its risk change rate, fluctuation value distribution, autocorrelation coefficient, texture characteristics, etc. are summarized.

[0088] In a specific embodiment, a risk identification model is constructed based on the annotated data features, including:

[0089] (1) If the data characteristics corresponding to any water conservancy project risk factor do not meet the preset threshold, the data to be analyzed corresponding to the data characteristics will be set as risk point data.

[0090] (2) Construct a machine learning model and input the data features corresponding to all key monitoring areas into the machine learning model for classification training. The machine learning model outputs recognition results based on the risk point data, where the recognition results include risk point category, risk point indicator, and risk point confidence.

[0091] (3) Adjust the parameters of the machine learning model based on the confidence level of the risk points, and set the adjusted machine learning model as the risk identification model.

[0092] Specifically, Figure 3A flowchart for the risk identification model construction and evaluation process is provided. Before constructing the risk identification model, the data characteristics need to be preliminarily screened to determine which data belongs to risk point data. For each water conservancy project risk factor (e.g., water level factor, seepage factor, structural deformation factor, etc.), a corresponding preset threshold is set, for example, for the water level factor, the threshold for the water level change rate is set to 0.5 meters / hour; for the seepage factor, the threshold for the seepage rate is set to 0.1 cubic meters / hour. If the data characteristics (such as risk change rate, fluctuation value distribution, etc.) corresponding to a certain water conservancy project risk factor do not meet the preset threshold, the data to be analyzed corresponding to the data characteristics are marked as risk point data, for example, if the water level change rate of a certain key monitoring area exceeds 0.5 meters / hour, the data corresponding to the water level change sequence in the key monitoring area are marked as risk point data.

[0093] A machine learning model is constructed using a random forest algorithm, and data characteristics corresponding to all key monitoring areas are input into the machine learning model for classification training. These data characteristics include physical characteristics, time series characteristics, and image characteristics, etc. During the training process, samples marked as risk point data are used as positive samples, and other normal data are used as negative samples to train the model to identify risk points. The identification results are output based on the input data characteristics. The risk point categories include, but are not limited to, water level anomalies, seepage anomalies, structural deformation anomalies, etc., the risk point indicators include, but are not limited to, water level change rate, seepage rate, displacement, etc., and the risk point confidence is the confidence of the machine learning model for the identification result, which is usually a value between 0 and 1, indicating the reliability of the machine learning model for the identification result.

[0094] The risk point confidence output by the machine learning model is analyzed. If the confidence is low (e.g., below a certain set threshold), it means that the model's identification of the risk point is not reliable enough. For the machine learning model corresponding to the random forest, the number of trees and the depth of the trees can be adjusted. The adjusted machine learning model is set as the final risk identification model for real-time monitoring of data risk identification.

[0095] In a specific embodiment, the risk prediction model outputs the risk factor change result of any key monitoring area within the prediction time, including:

[0096] (1) Based on data correlation, the risk identification data corresponding to any key monitoring area are combined in sequence to generate multiple risk sequence groups, and the risk sequence groups contained in all key monitoring areas are preprocessed and classified into a training set and a test set.

[0097] (2) A risk prediction model is constructed using a deep learning framework, the training set is input into the risk prediction model for training, and the test set is used to evaluate the trained risk prediction model to calculate the prediction error.

[0098] (3) Set the risk category label of the risk sequence group based on the sensor type, input the risk sequence group contained in any key monitoring area into the risk prediction model, output the change sequence in the prediction time based on the risk category label, correct the change sequence based on the prediction error and set it as the risk factor change result.

[0099] Specifically, the correlation between the risk identification data is analyzed, and the data correlation can be calculated by the Pearson correlation coefficient, for example, the relationship between water level change and seepage rate, the relationship between structure deformation and rainfall, etc. The data with correlation is combined into multiple risk sequence groups. For example, the water level change data and the seepage rate data are combined into a risk sequence group, and the structure deformation data and the rainfall data are combined into another risk sequence group. Preprocessing refers to data processing procedures such as data cleaning and normalization processing. Generally, the training set is used for model training, and the test set is used for model evaluation, for example, 70% of the data corresponding to the risk sequence group can be used as the training set, and 30% of the data corresponding to the risk sequence group can be used as the test set.

[0100] A deep learning framework suitable for time series prediction is selected, for example, a combination of convolutional neural network (CNN) and LSTM (CNN-LSTM), etc. The model parameters are adjusted by the back propagation algorithm to optimize the model performance, for example, adjusting the learning rate, the number of hidden layer nodes, the regularization parameter, etc. The risk sequence group in the test set is input into the trained risk prediction model to evaluate the prediction performance of the model. The error between the predicted value and the actual value is calculated, which is the prediction error. Common error indicators include mean square error (MSE), mean absolute error (MAE), etc.

[0101] According to the sensor type, a risk category label is set for each risk sequence group. For example, the risk category label corresponding to the water level sensor is "water level anomaly", and the risk category label corresponding to the seepage sensor is "seepage anomaly". The risk prediction model outputs the change sequence in the prediction time based on the risk category label, for example, outputs the change sequence of water level in the next 24 hours, the change sequence of seepage rate, etc. The change sequence is corrected according to the prediction error. If the prediction error is large, the accuracy of the prediction can be improved by adjusting the model parameters or introducing an error correction mechanism.

[0102] In a specific embodiment, if there is a discrete point greater than the first preset value in the risk factor change result, any key monitoring area is set as a first water conservancy risk point, and the risk category label corresponding to the discrete point is set as the risk factor.

[0103] Specifically, the discrete points usually indicate potential burst risks, such as dam seepage, rapid rise of water level, etc. By detecting the discrete points, these abnormal situations can be found in time, and the risk is further expanded. If the discrete point is detected, the key monitoring area is determined as the first water conservancy risk point, and the specific risk factors of the risk point are determined, so as to facilitate the water conservancy engineering management personnel to quickly locate the problem and take targeted measures.

[0104] In a specific embodiment, the second water conservancy risk point is screened out from all key monitoring areas based on the causal relationship strength, comprising:

[0105] (1) Set any key monitoring area as a regional node, set a directed edge based on the operation layout between nodes, connect all regional nodes using the directed edge, and generate a causal relationship graph.

[0106] (2) Split the causal relationship graph into multiple sub-relationship graphs based on the directed connection mode, if there is a regional node corresponding to the first water conservancy risk point in the sub-relationship graph, then based on the sub-relationship graph, the Bayesian network structure is established, and the conditional probability distribution of the Bayesian network structure is estimated using the real-time monitoring data of the regional node in the sub-relationship graph, and the causal relationship strength is generated.

[0107] (3) If the causal relationship strength is greater than the second preset value, set the regional node of the first water conservancy risk point in the sub-relationship graph as the second water conservancy risk point.

[0108] Specifically, each key monitoring area is set as a regional node. For example, the dam body, dam foundation, spillway, sluice, etc. of the dam are respectively taken as different regional nodes. The directed edge represents the influence direction of one regional node to another regional node. For example, the increase of rainfall (regional node A) may cause the rise of reservoir water level (regional node B), and a directed edge is drawn from regional node A to regional node B. All regional nodes are connected using directed edges to generate a complete causal relationship graph. The graph reflects the causal relationship between each key monitoring area in the water conservancy project.

[0109] The causal relationship graph is split into multiple sub-relationship graphs, each of which contains a group of interrelated regional nodes and directed edges. For example, a sub-relationship graph may contain rainfall nodes, reservoir water level nodes and dam seepage nodes of different regional nodes. If a sub-relationship graph contains a first water conservancy risk point (such as a region has been determined to be high risk), further analysis is performed on the sub-relationship graph.

[0110] ​​​​​​​A Bayesian network structure is established using the regional nodes and directed edges in the sub-relationship graph. A Bayesian network is a graphical representation method based on a probability model that is used to describe the conditional dependency relationship between variables. Using the real-time monitoring data corresponding to the regional nodes in the sub-relationship graph, the conditional probability distribution of the Bayesian network structure is estimated. For example, the probability of a reservoir water level rising when rainfall increases is calculated. The numerical value corresponding to the conditional probability is set as the causal relationship strength, which is used to represent the degree of influence of one regional node on another regional node. For example, the influence of rainfall on reservoir water level and the influence of reservoir water level on dam seepage are calculated.

[0111] Based on the specific circumstances and experience of water conservancy projects, a threshold (second preset value) is set to determine whether the causal relationship strength is sufficiently strong. For example, the second preset value is set to 0.7, indicating that if the causal relationship strength is greater than 0.7, the impact is considered significant. If the causal relationship strength in a sub-relationship graph is greater than the second preset value, the regional subnode (i.e., the affected regional node) of the first water conservancy risk point in the sub-relationship graph is set as the second water conservancy risk point. For example, if the causal relationship strength of increased rainfall leading to a rise in reservoir water level is greater than 0.7, the reservoir water level node is set as the second water conservancy risk point.

[0112] The above describes the method for intelligent identification of water conservancy project risk points based on the Internet of Things in the embodiment of the present application. The following describes the intelligent identification system for water conservancy project risk points based on the Internet of Things in the embodiment of the present application. Figure 4 In the embodiment of the present application, an embodiment of the intelligent identification system for water conservancy project risk points based on the Internet of Things includes:

[0113] The data transmission module 201 is used to obtain the regional distribution map of the water conservancy project, and deploy multiple sensors in the key monitoring areas of the water conservancy project based on the regional distribution map. The sensors collect sensor monitoring data based on a preset collection method and transmit the sensor monitoring data to the Internet of Things gateway. The Internet of Things gateway pre-processes the received sensor monitoring data and classifies the sensor monitoring data based on the key monitoring area and sensor type and stores it in the cloud server.

[0114] The risk identification module 202 is used to set the sensor monitoring data contained in any key monitoring area as the data to be analyzed, extract data features related to the risk factors of water conservancy projects from the data to be analyzed, mark the existing risk point data based on the data features, build a risk identification model based on the marked data features, and deploy the risk identification model to the cloud server.

[0115] The risk prediction module 203 is used for the cloud server to receive the real-time monitoring data of any key monitoring area. The risk identification model outputs the risk identification data based on the real-time monitoring data, performs time series modeling on all risk identification data, and generates a risk prediction model. The risk prediction model outputs the risk factor change results of any key monitoring area within the prediction time, and screens out the first water conservancy risk point in all key monitoring areas based on the risk factor change results.

[0116] The summary module 204 is used to construct a causal relationship diagram for all key monitoring areas based on the regional distribution map, quantify the causal relationship strength in the causal relationship diagram based on the correlation of water conservancy project risk factors, screen out the second water conservancy risk point in all key monitoring areas based on the causal relationship strength, summarize the first water conservancy risk point and the second water conservancy risk point based on the water conservancy project risk factors, and generate a risk point list.

[0117] Through the collaborative efforts of the aforementioned components, the system first utilizes IoT technology to deploy multiple sensors in key monitoring areas of water conservancy projects. The collected monitoring data is preprocessed by an IoT gateway and stored in a classified manner on a cloud server. Through the interaction of route request and response information, sensors can quickly find the optimal path to the IoT gateway. Collaboration between adjacent nodes and the dynamic updating of label information further optimize the path selection process, ensuring network stability and data transmission accuracy. Data features are then extracted from three dimensions: physical, time series, and image models. These features comprehensively reflect the dynamic changes, temporal correlations, and spatial distribution of water conservancy project risk factors. By calculating the risk change rate and fluctuation value distribution, the rate of change and fluctuation of risk factors are quantified, providing an important physical basis for risk identification. Extracted autocorrelation coefficients, partial autocorrelation coefficients, and wavelet transform features reflect the long-term trends and cyclical changes of risk factors, facilitating risk prediction. Texture features, edge detection features, and target monitoring features intuitively reflect the spatial distribution and morphological changes of risk factors. Different types of features are aggregated according to the categories of water conservancy project risk factors, forming a comprehensive data feature set that provides rich input for the subsequent construction of risk identification models. Finally, a suitable machine learning model is selected to classify and identify risk points in water conservancy projects. It can learn the complex relationship between data features and risk points, realize automatic identification and classification of risk points, and use a deep learning framework to build a risk prediction model. It can learn the complex patterns and dynamic changes of risk sequence groups, improve the accuracy of risk prediction, and realize intelligent prediction of risk factors through deep learning models, reduce manual intervention, and improve the automation level of water conservancy project risk monitoring.

[0118] This application also constructs a causal relationship diagram to intuitively represent the mutual influence relationship between key monitoring areas, establishes a Bayesian network structure based on the sub-relationship diagram, and estimates the conditional probability distribution, which can quantify the causal relationship between variables, improve the scientificity and accuracy of risk point screening, calculate the causal relationship strength through the Bayesian network, quantify the degree of mutual influence between nodes in each area, and provide a clear quantitative standard for screening the second water conservancy risk point. The dynamic calculation of the causal relationship strength through the Bayesian network and real-time monitoring data can adapt to the dynamic changes in the operation of water conservancy projects and improve the intelligence level of risk monitoring.

[0119] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0120] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program code.

[0121] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An intelligent identification method for water conservancy project risk points based on the Internet of Things, characterized by: The method for intelligently identifying risk points in water conservancy projects based on the Internet of Things includes: Obtaining a regional distribution map of the water conservancy project, deploying multiple sensors in key monitoring areas of the water conservancy project based on the regional distribution map, wherein the sensors collect sensor monitoring data based on a preset collection method and transmit the sensor monitoring data to an Internet of Things gateway, wherein the Internet of Things gateway pre-processes the received sensor monitoring data and classifies and stores the sensor monitoring data in a cloud server based on the key monitoring areas and sensor types; The sensor monitoring data contained in any key monitoring area is set as data to be analyzed, data features related to water conservancy project risk factors are extracted from the data to be analyzed, existing risk point data are labeled based on the data features, a risk identification model is constructed based on the labeled data features, and the risk identification model is deployed to the cloud server; The cloud server receives real-time monitoring data of any of the key monitoring areas, the risk identification model outputs risk identification data based on the real-time monitoring data, performs time series modeling on all the risk identification data to generate a risk prediction model, the risk prediction model outputs the risk factor change results of any of the key monitoring areas within the prediction time, and screens out the first water conservancy risk point in all the key monitoring areas based on the risk factor change results; constructing a causal relationship diagram for all the key monitoring areas based on the regional distribution map, quantifying the causal relationship strength in the causal relationship diagram based on the correlation of the water conservancy project risk factors, screening out second water conservancy risk points in all the key monitoring areas based on the causal relationship strength, and summarizing the first water conservancy risk points and the second water conservancy risk points based on the water conservancy project risk factors to generate a risk point list; The step of extracting data features related to water conservancy project risk factors from the data to be analyzed includes: Based on the water conservancy project risk factors, risk factor data are extracted from the data to be analyzed, wherein the water conservancy project risk factors include water level factors, leakage factors, structural deformation factors and flow rate factors; a physical model is established based on the sensor type, the risk factor data is standardized and then output to the physical model, the ratio of the difference between the risk factor data at adjacent time points and the time interval is calculated, the ratio is set as the risk change rate, the difference between the maximum and minimum values ​​of the risk factor data is calculated within a preset time window, all the differences are summarized and set as a fluctuation value distribution, the risk change rate and the The fluctuation value distribution is set as a physical feature; a sequence analysis model is established based on the data type, and the sequence analysis model extracts time series features from the risk factor data, and the time series features include autocorrelation coefficients, partial autocorrelation coefficients, and wavelet transform features; an image analysis model is established, and the risk factor data is input into the image analysis model based on the data type to extract image features, and the image features include texture features, edge detection features, and target monitoring features; the physical features, the time series features, and the image features are summarized based on the category of the water conservancy project risk factors and set as the data features; The step of constructing a risk identification model based on the labeled data features includes: If the data feature corresponding to any of the water conservancy project risk factors does not meet the preset threshold, the data to be analyzed corresponding to the data feature is set as the risk point data; a machine learning model is constructed, and the data features corresponding to all the key monitoring areas are input into the machine learning model for classification training, and the machine learning model outputs an identification result based on the risk point data, wherein the identification result includes the risk point category, risk point index, and risk point confidence; the parameters of the machine learning model are adjusted based on the risk point confidence, and the adjusted machine learning model is set as the risk identification model; The risk prediction model outputs the risk factor change results of any key monitoring area within the prediction time, including: Based on data correlation, the risk identification data corresponding to any of the key monitoring areas are serially combined to generate multiple risk sequence groups, and the risk sequence groups contained in all the key monitoring areas are pre-processed and classified into training sets and test sets; the risk prediction model is constructed using a deep learning framework, the training set is input into the risk prediction model for training, the trained risk prediction model is evaluated using the test set, and the prediction error is calculated; the risk category label of the risk sequence group is set based on the sensor type, the risk sequence group contained in any of the key monitoring areas is input into the risk prediction model, the change sequence within the prediction time is output based on the risk category label, the change sequence is corrected based on the prediction error and set as the risk factor change result.

2. The method for intelligently identifying risk points of water conservancy projects based on the Internet of Things according to claim 1 is characterized in that: The transmitting the sensor monitoring data to the Internet of Things gateway includes: Connecting all the sensors and the IoT gateway to the same IoT network, configuring communication parameters for the sensors, and setting unique node IDs for all the sensors and the IoT gateway; When collecting and generating the sensor monitoring data, the sensor checks whether there is valid routing information to reach the Internet of Things gateway. If so, the sensor monitoring data is transmitted to the Internet of Things gateway based on the arrival path of the valid routing information. If not, an optimal path is generated in the Internet of Things network, and the sensor monitoring data is transmitted to the Internet of Things gateway based on the optimal path.

3. The method for intelligently identifying risk points of water conservancy projects based on the Internet of Things according to claim 2 is characterized in that: Generating an optimal path in the Internet of Things network includes: The sensor transmits the routing request information to all neighboring nodes; After receiving the routing request information, the adjacent node checks whether the routing request information has passed through the current node. If the current node is the destination node corresponding to the IoT gateway, the response processing phase is entered. Otherwise, the ID tag of the current node is added to the request tag corresponding to the routing request information, and the routing request information is transmitted to the remaining adjacent nodes. After receiving the routing request information, the destination node generates routing response information and adds its own node ID to the routing response information. The routing response information is returned to the sensor in the opposite direction of the routing request information. After receiving the routing response information, the intermediate node extracts the label information, generates an extended label, adds the extended label to the routing response information, and continues to send the routing response information to the sensor. The intermediate node uses the received label information as the basic label and generates a new extended label for the next node; After receiving a plurality of the routing response messages, the sensor sets the path represented by the longest tag information as the optimal path based on the tag length.

4. The method for intelligently identifying risk points of water conservancy projects based on the Internet of Things according to claim 1 is characterized in that: If there are discrete points greater than the first preset value in the risk factor change results, any of the key monitoring areas is set as the first water conservancy risk point, and the risk category label corresponding to the discrete point is set as the risk factor.

5. The method for intelligently identifying risk points of water conservancy projects based on the Internet of Things according to claim 1 is characterized in that: The step of screening out the second water conservancy risk point in all the key monitoring areas based on the strength of the causal relationship includes: Setting any of the key monitoring areas as a regional node, setting directed edges based on the operation layout between the nodes, and connecting all the regional nodes using the directed edges to generate the causal relationship graph; Splitting the causal relationship graph into multiple sub-graphs based on a directed connection method, if the regional node corresponding to the first water conservancy risk point exists in the sub-graph, establishing a Bayesian network structure based on the sub-graph, and estimating the conditional probability distribution of the Bayesian network structure using the real-time monitoring data corresponding to the regional node in the sub-graph to generate the causal relationship strength; If the causal relationship strength is greater than a second preset value, the regional sub-node of the first water risk point in the sub-relationship graph is set as the second water risk point.

6. An intelligent identification system for water conservancy project risk points based on the Internet of Things, characterized by: The IoT-based intelligent identification system for water conservancy project risk points includes: a data transmission module, configured to obtain a regional distribution map of the water conservancy project, deploy a plurality of sensors in key monitoring areas of the water conservancy project based on the regional distribution map, wherein the sensors collect sensor monitoring data based on a preset collection method and transmit the sensor monitoring data to an Internet of Things gateway, wherein the Internet of Things gateway pre-processes the received sensor monitoring data and classifies and stores the sensor monitoring data in a cloud server based on the key monitoring areas and sensor types; a risk identification module, configured to set the sensor monitoring data contained in any key monitoring area as data to be analyzed, extract data features related to water conservancy project risk factors from the data to be analyzed, annotate existing risk point data based on the data features, construct a risk identification model based on the annotated data features, and deploy the risk identification model to the cloud server; A risk prediction module is configured to receive real-time monitoring data of any of the key monitoring areas from the cloud server, output risk identification data based on the real-time monitoring data from the risk identification model, perform time series modeling on all of the risk identification data to generate a risk prediction model, output a risk factor change result for any of the key monitoring areas within a prediction time, and screen out a first water conservancy risk point from all of the key monitoring areas based on the risk factor change result; a summarization module, configured to construct a causal relationship diagram for all the key monitoring areas based on the regional distribution map, quantify the causal relationship strength in the causal relationship diagram based on the correlation of the water conservancy project risk factors, screen out second water conservancy risk points in all the key monitoring areas based on the causal relationship strength, and summarize the first water conservancy risk points and the second water conservancy risk points based on the water conservancy project risk factors to generate a risk point list; The step of extracting data features related to water conservancy project risk factors from the data to be analyzed includes: Based on the water conservancy project risk factors, risk factor data are extracted from the data to be analyzed, wherein the water conservancy project risk factors include water level factors, leakage factors, structural deformation factors and flow rate factors; a physical model is established based on the sensor type, the risk factor data is standardized and then output to the physical model, the ratio of the difference between the risk factor data at adjacent time points and the time interval is calculated, the ratio is set as the risk change rate, the difference between the maximum and minimum values ​​of the risk factor data is calculated within a preset time window, all the differences are summarized and set as a fluctuation value distribution, the risk change rate and the The fluctuation value distribution is set as a physical feature; a sequence analysis model is established based on the data type, and the sequence analysis model extracts time series features from the risk factor data, and the time series features include autocorrelation coefficients, partial autocorrelation coefficients, and wavelet transform features; an image analysis model is established, and the risk factor data is input into the image analysis model based on the data type to extract image features, and the image features include texture features, edge detection features, and target monitoring features; the physical features, the time series features, and the image features are summarized based on the category of the water conservancy project risk factors and set as the data features; The step of constructing a risk identification model based on the labeled data features includes: If the data feature corresponding to any of the water conservancy project risk factors does not meet the preset threshold, the data to be analyzed corresponding to the data feature is set as the risk point data; a machine learning model is constructed, and the data features corresponding to all the key monitoring areas are input into the machine learning model for classification training, and the machine learning model outputs an identification result based on the risk point data, wherein the identification result includes the risk point category, risk point index, and risk point confidence; the parameters of the machine learning model are adjusted based on the risk point confidence, and the adjusted machine learning model is set as the risk identification model; The risk prediction model outputs the risk factor change results of any key monitoring area within the prediction time, including: Based on data correlation, the risk identification data corresponding to any of the key monitoring areas are serially combined to generate multiple risk sequence groups, and the risk sequence groups contained in all the key monitoring areas are pre-processed and classified into training sets and test sets; the risk prediction model is constructed using a deep learning framework, the training set is input into the risk prediction model for training, the trained risk prediction model is evaluated using the test set, and the prediction error is calculated; the risk category label of the risk sequence group is set based on the sensor type, the risk sequence group contained in any of the key monitoring areas is input into the risk prediction model, the change sequence within the prediction time is output based on the risk category label, the change sequence is corrected based on the prediction error and set as the risk factor change result.

Images