Intelligent water conservancy information management methods and platforms

Abstract

This application provides an intelligent water conservancy information management method and platform, belonging to the field of intelligent management technology. The method includes: acquiring hydrological data, equipment status data, and equipment health data; constructing a basin-level map based on the hydrological data, equipment status data, and equipment health data; wherein the basin-level map includes multiple equipment-level subgraphs; constructing a physical consistency regularization term based on preset data-driven loss and preset physical consistency loss; inputting the basin-level map, equipment-level subgraphs, and physical consistency regularization term into a preset graph neural network fusion model to obtain a scheduling strategy for each edge device; and sending the scheduling strategy to each corresponding edge device. This achieves collaborative optimization scheduling across basins and among multiple devices.

Description

Technical Field

[0001] This application relates to the field of intelligent management technology, and in particular to an intelligent water conservancy information management method and platform. Background Technology

[0002] Water conservancy information management, as an important means to ensure the sustainable use of water resources and improve the operational efficiency of water projects, has received widespread attention. Currently, water conservancy information management is widely applied in various aspects such as basin water resource scheduling, hydraulic structure operation management, and aquatic ecological environment monitoring. Among related technologies, most water conservancy information management platforms adopt a "unit-based" design, treating reservoirs, pumping stations, gates, and other equipment as independent nodes for management, making it difficult to achieve coordinated and optimized scheduling across basins and among multiple devices. Therefore, how to enable water conservancy information management to achieve coordinated and optimized scheduling across basins and among multiple devices has become an urgent technical problem to be solved. Summary of the Invention

[0003] The main objective of this application is to propose an intelligent water conservancy information management method and platform, which aims to enable water conservancy information management to achieve collaborative optimization scheduling across river basins and multiple devices.

[0004] To achieve the above objectives, a first aspect of this application proposes an intelligent water conservancy information management method applied to a cloud server, the method comprising:

[0005] Acquire hydrological data, equipment status data, and equipment health data;

[0006] A watershed-level map is constructed based on the hydrological data, the equipment status data, and the equipment health data; wherein, the watershed-level map includes multiple equipment-level sub-maps;

[0007] Based on the preset data-driven loss and the preset physical consistency loss, a physical consistency regularization term is constructed;

[0008] The watershed-level graph, the device-level subgraph, and the physical consistency regularization term are input into a preset graph neural network fusion model to obtain a scheduling strategy for each edge device; wherein, the scheduling strategy is used to indicate the operating status of the edge device;

[0009] The scheduling policy is sent to each of the corresponding edge devices.

[0010] In some embodiments, after sending the scheduling policy to each of the corresponding edge devices, the method further includes:

[0011] The system receives risk score data and the optimized scheduling strategy sent by the edge device; wherein the optimized scheduling strategy is obtained by the edge device after modifying the scheduling strategy according to the dual-sensitivity optimization model, and the risk score data is obtained by the edge-level health risk model to assess the risk of the scheduling strategy.

[0012] The preset joint feedback optimization model is trained based on the scheduling strategy, the optimized scheduling strategy, the equipment health data before the scheduling strategy is executed, the equipment health data after the scheduling strategy is executed, and the risk score data to obtain the strategy residual modeling loss and the state prediction loss.

[0013] The graph neural network fusion model is updated with parameters based on the strategy residual modeling loss to obtain the updated graph neural network fusion model.

[0014] The parameters of the edge-level health risk model and the dual-sensitivity optimization model are updated based on the state prediction loss to obtain the updated edge-level health risk model and the updated dual-sensitivity optimization model.

[0015] In some embodiments, constructing a physical consistency regularization term based on a preset data-driven loss and a preset physical consistency loss includes:

[0016] Obtain actual observation data;

[0017] The data-driven loss is constructed based on the error between the scheduling strategy and the actual observed data.

[0018] To ensure that the scheduling strategy conforms to the physical constraints of the edge device, the physical consistency loss is constructed.

[0019] The physical consistency regularization term is constructed based on the data-driven loss and the physical consistency loss.

[0020] To achieve the above objectives, a second aspect of this application proposes an intelligent water conservancy information management method applied to edge devices, the method comprising:

[0021] The system receives a scheduling policy sent by a cloud server; wherein the scheduling policy is obtained by the cloud server based on a watershed-level graph, a device-level subgraph, and a physical consistency regularization term.

[0022] The scheduling strategy is risk-assessed based on a preset edge-level health risk model to obtain risk score data.

[0023] If the risk score data is less than a preset first risk threshold, the scheduling strategy is executed. If the risk score data is greater than or equal to the first risk threshold and less than a preset second risk threshold, the amplitude is limited before the scheduling strategy is executed. If the risk score data is greater than or equal to the second risk threshold, an alarm message is sent.

[0024] In some embodiments, the method further includes: constructing the edge-level health risk model, specifically including:

[0025] Based on the device-level subgraph and the scheduling strategy, construct a structure-aware regularization term;

[0026] The edge-level health risk model is constructed based on the current device health data, the scheduling strategy, the preset feature cross-mapping function, and the structure-aware regularization term.

[0027] In some embodiments, if the risk score data is greater than or equal to the second risk threshold, the method further includes:

[0028] A dual-sensitivity optimization model is constructed based on the scheduling strategy, the risk scoring data, the current device health data, and the structure-aware regularization term.

[0029] The scheduling strategy is modified according to the dual-sensitivity optimization model to obtain the optimized scheduling strategy.

[0030] To achieve the above objectives, a third aspect of this application proposes an intelligent water conservancy information management platform applied to a cloud server, the management platform comprising:

[0031] The acquisition module is used to acquire hydrological data, equipment status data, and equipment health data.

[0032] The first construction module is used to construct a watershed-level map based on the hydrological data, the equipment status data, and the equipment health data; wherein the watershed-level map includes multiple equipment-level sub-maps;

[0033] The second construction module is used to construct a physical consistency regularization term based on the preset data-driven loss and the preset physical consistency loss.

[0034] The input module is used to input the watershed-level graph, the device-level subgraph, and the physical consistency regularization term into a preset graph neural network fusion model to obtain the scheduling strategy for each edge device; wherein, the scheduling strategy is used to indicate the operating status of the edge device;

[0035] The sending module is used to send the scheduling policy to each of the corresponding edge devices.

[0036] To achieve the above objectives, a fourth aspect of this application proposes an intelligent water conservancy information management platform, applied to edge devices, the management platform comprising:

[0037] The receiving module is used to receive the scheduling policy sent by the cloud server; wherein the scheduling policy is obtained by the cloud server based on the watershed level graph, the device level subgraph, and the physical consistency regularization term;

[0038] The evaluation module is used to evaluate the scheduling strategy based on a preset edge-level health risk model and obtain risk score data.

[0039] The execution module is configured to execute the scheduling strategy if the risk score data is less than a preset first risk threshold, perform amplitude limiting processing before executing the scheduling strategy if the risk score data is greater than or equal to the first risk threshold and less than a preset second risk threshold, and send an alarm message if the risk score data is greater than or equal to the second risk threshold.

[0040] To achieve the above objectives, a fifth aspect of the present application provides an electronic device, the electronic device including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the methods described in the first aspect and the methods described in the second aspect.

[0041] To achieve the above objectives, a sixth aspect of the present application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the methods described in the first aspect and the methods described in the second aspect.

[0042] The intelligent water conservancy information management method and platform proposed in this application acquires hydrological data, equipment status data, and equipment health data. A basin-level map is constructed based on these data, including multiple equipment-level subgraphs. A physical consistency regularization term is constructed based on preset data-driven losses and preset physical consistency losses. The basin-level map, equipment-level subgraphs, and physical consistency regularization term are input into a preset graph neural network fusion model to obtain a scheduling strategy for each edge device. The scheduling strategy is then sent to each corresponding edge device, thereby achieving collaborative optimization scheduling across basins and among multiple devices. Attached Figure Description

[0043] Figure 1 This is a flowchart of the intelligent water conservancy information management method provided in the embodiments of this application;

[0044] Figure 2 yes Figure 1The flowchart of step S103 in the process;

[0045] Figure 3 This is a flowchart of an intelligent water conservancy information management method provided in another embodiment of this application;

[0046] Figure 4 This is a flowchart of the intelligent water conservancy information management method provided in the third embodiment of this application;

[0047] Figure 5 This is a flowchart of the intelligent water conservancy information management method provided in the fourth embodiment of this application;

[0048] Figure 6 This is a schematic diagram of the structure of the intelligent water conservancy information management platform provided in the embodiments of this application;

[0049] Figure 7 This is a schematic diagram of the structure of an intelligent water conservancy information management platform provided in another embodiment of this application;

[0050] Figure 8 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0052] It should be noted that although functional modules are divided in the system diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the system or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, and the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0053] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0054] Water conservancy information management, as an important means to ensure the sustainable use of water resources and improve the operational efficiency of water projects, has received widespread attention. Currently, water conservancy information management platforms are widely used in various aspects such as river basin water resource scheduling, hydraulic structure operation management, and aquatic ecological environment monitoring. However, most related technologies in water conservancy information management adopt a "unit-based" design, treating reservoirs, pumping stations, and gates as independent nodes for management, making it difficult to achieve coordinated and optimized scheduling across river basins and among multiple devices. Therefore, how to enable coordinated and optimized scheduling across river basins and among multiple devices through water conservancy information management has become an urgent technical problem to be solved.

[0055] Based on this, this application provides an intelligent water conservancy information management method and platform, aiming to construct a basin-level map based on hydrological data, equipment status data, and equipment health data of all edge devices. The basin-level map includes multiple equipment-level subgraphs. The basin-level map, equipment-level subgraphs, and physical consistency regularization terms are then input into a preset graph neural network fusion model to obtain a scheduling strategy for each edge device, and the scheduling strategy is sent to each edge device. Edge devices include reservoirs, pumping stations, and gates, thereby achieving collaborative optimization scheduling across basins and among multiple devices.

[0056] The intelligent water conservancy information management method and platform provided in this application are specifically described through the following embodiments. First, the intelligent water conservancy information management method in this application embodiment is described.

[0057] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0058] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0059] The intelligent water conservancy information management method provided in this application can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, etc.; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the intelligent water conservancy information management method, but is not limited to the above forms.

[0060] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0061] Please refer to Figure 1 , Figure 1 This is a flowchart of the intelligent water conservancy information management method provided in the embodiments of this application, which is applied to a cloud server. Figure 1 The method may include, but is not limited to, steps S101 to S105.

[0062] Step S101: Obtain hydrological data, equipment status data, and equipment health data;

[0063] Step S102: Construct a watershed-level map based on hydrological data, equipment status data, and equipment health data; wherein, the watershed-level map includes multiple equipment-level sub-maps;

[0064] Step S103: Construct a physical consistency regularization term based on the preset data-driven loss and the preset physical consistency loss;

[0065] Step S104: Input the watershed-level graph, device-level subgraph, and physical consistency regularization term into the preset graph neural network fusion model to obtain the scheduling strategy for each edge device; wherein, the scheduling strategy is used to indicate the operating status of the edge device;

[0066] Step S105: Send the scheduling policy to each corresponding edge device.

[0067] In step S101 of some embodiments, the hydrological data comes from equipment such as hydrological stations, meteorological stations, and remote sensing satellites within the watershed, including flow rate. Water level Rainfall, etc. Assuming the data is obtained from the flow meter The system records changes in reservoir flow over the past 24 hours; this data is typically recorded in time-series format, ensuring high timeliness. Equipment status data comes from internal sensors and the Supervisory Control and Data Acquisition (SCADA) system. It monitors the opening and closing status of pump stations and gates, pump load, etc. For example, pump station load data... It will record the real-time load of the pump set and the opening data of the gate. It will reflect its open / closed status.

[0068] Equipment health data comes from internal sensors such as stress sensors and vibration sensors, reflecting the equipment's health status. For example, stress data from a dam body. Vibration data of pump units These data are crucial for the stable operation of equipment, especially in the management of large-scale equipment such as reservoirs and sluice gates.

[0069] In step S102 of some embodiments, the collected hydrological data, equipment status data, and equipment health data are mapped to a watershed-level map. In the diagram. This represents various facilities within the watershed (such as reservoirs, pumping stations, and sluice gate groups), while the edges... This indicates the dependency between equipment, such as the impact of changes in upstream reservoir flow on downstream pumping stations.

[0070] The dependencies between devices are represented in the graph by the edge weights. To reflect this. For example, edge weights. The correlation between reservoir flow and pump station flow can be used to calculate the degree of hydrological dependence. The formula for calculating the weight is shown in formula (1) below: (1)

[0071] in, Represents the weight of the edge. This represents the reservoir's flow data. Indicates pumping station The flow data allows for the quantification of the interactions between devices within the watershed.

[0072] It should be noted that the watershed-level map Each device node is further subdivided into a device-level subgraph. This subgraph contains multiple units within the equipment and their control relationships. For example, a single pumping station can contain multiple pump groups, and the operating status of each pump group affects the overall operating efficiency of the pumping station. Therefore, these pump groups need to be treated as subgraph nodes.

[0073] Each node in the device-level subgraph Representing a single device (such as a pump set, gate, etc.), side Describe the collaborative control relationships between devices. For example, multiple pump sets may have a requirement for synchronized start-stop; this collaborative relationship would be represented by edge weights.

[0074] Each device node Features include not only its spatial information (such as location coordinates) This also includes the device's status data, such as load. and equipment health data (such as dam stress) ,vibration These features will become key inputs for prediction and scheduling in subsequent modeling.

[0075] On the edges of the graph and In this process, the hydrological dependence and control dependence between equipment are integrated to ensure that subsequent scheduling decisions fully consider the coupling relationships between equipment. For example: equipment Status (such as traffic) ) for equipment The operation of (such as pumping stations) has an impact. This reflects the intensity of this impact. In coordinated control between devices, such as between multiple pump sets, the side... Weights in These indicate the degree of control dependency between them, such as the frequency of coordinated start-stop and the strength of synchronization control. These edge weights will serve as one of the most fundamental control information in subsequent intelligent scheduling, used to determine how different devices should work together, thereby avoiding system conflicts or resource waste.

[0076] In practical applications, to achieve efficient management of watershed-level graphs and device-level subgraphs, a nested graph index table is created. It will be a watershed-level node Its corresponding device-level subgraph Mapping is performed. This ensures that subsequent steps can quickly access and process the status of specific devices based on the watershed-level graph information. For example, suppose there are nodes in the watershed-level graph. (Representing Reservoir 1), its corresponding equipment-level sub-diagram It includes multiple pumping stations and gates. (Through...) It allows for quick retrieval and updating of data for these sub-devices.

[0077] This step effectively integrates this data and transforms it into a unified graph model, enabling better intelligent scheduling and decision-making across watersheds and equipment. Integrating heterogeneous data from different sources into a unified multi-scale graph structure provides a solid foundation for subsequent scheduling, execution, and feedback. The hierarchical design of the graph allows the system to simultaneously capture global dependencies at the watershed level and local relationships at the equipment level, providing physically consistent input for subsequent graph neural network fusion models. This process not only ensures the physical consistency of the data but also provides sufficient upstream and downstream dependency information for subsequent scheduling decisions. Ultimately, this nested graph structure can flexibly respond to changes in water conservancy equipment (such as the addition or decommissioning of equipment) and provide strong data support for subsequent models.

[0078] Please see Figure 2 In some embodiments, step S103 may include steps S201 to S204:

[0079] Step S201: Obtain actual observation data;

[0080] Step S202: Construct a data-driven loss based on the error between the scheduling strategy and the actual observation data;

[0081] Step S203: Make the scheduling strategy conform to the physical constraints of the edge devices and construct the physical consistency loss.

[0082] Step S204: Construct a physical consistency regularization term based on the data-driven loss and the physical consistency loss.

[0083] In steps S201 to S202 of some embodiments, to ensure that the scheduling strategy conforms to physical constraints, a physical consistency regularization term needs to be added to the loss function of the graph neural network fusion model. The physical consistency regularization term forces the graph neural network fusion model to consider the physical limitations of equipment operation, such as maximum safe values ​​for water level and flow rate, when generating the scheduling strategy. The physical consistency regularization term includes data-driven loss and physical consistency loss, where the data-driven loss represents the error between the scheduling strategy generated by the model and the actual observed data. Therefore, it is necessary to obtain the actual observed data and construct the data-driven loss based on the error between the scheduling strategy and the actual observed data.

[0084] In steps S203 to S204 of some embodiments, the physical consistency loss is used to ensure that the scheduling strategy output by the graph neural fusion model conforms to the physical constraints of the device, preventing the scheduling strategy from violating limits such as water level and flow rate. For example, the upper limit constraint of water level can be expressed as shown in the following formula (2): (2)

[0085] in, Indicates the loss of physical consistency. Reservoir The actual water level Reservoir The maximum water level. If the pump station's start-up and shutdown strategy causes the water level to exceed the threshold, the loss of physical consistency will force the graph neural network fusion model to readjust its scheduling strategy to avoid unsafe scheduling.

[0086] The physical consistency regularization term is shown in the following formula (3): (3)

[0087] in, This represents the physical consistency regularization term. Indicates data-driven loss. Indicates the loss of physical consistency. This represents the weighting coefficient of physical consistency loss in the total loss function, used to control the degree of influence of physical constraints on model optimization. It can be initially set to 0.3.

[0088] Steps S201 to S204, as illustrated in the embodiments of this application, construct a physical consistency regularization term using data-driven loss and physical consistency loss. Adding a physical consistency regularization term to the loss function of the graph neural network fusion model forces the model to consider the physical limitations of device operation when generating scheduling strategies, thereby improving the safety of device operation.

[0089] In steps S104 to S105 of some embodiments, the graph neural fusion model (GNN) is used to fuse the input watershed-level graph, device-level subgraph, and physical consistency regularization term to generate a scheduling strategy for each device through graph convolution operations. The graph convolution operation at each layer combines the dependencies between devices and the constraints of the physical consistency regularization term to ensure that the scheduling strategy is not only based on historical data but also conforms to physical constraints. Among them, the graph convolution operation includes each device node updating according to the information of its neighboring nodes. The update process is based not only on the features of the neighboring nodes but also on the constraints of the physical consistency regularization term on the model. The graph convolution formula is shown in the following formula (4): (4)

[0090] in, Indicates the first Layer nodes Feature representation, Indicates the first Layer nodes The characteristics are represented. Represents a node The set of adjacent nodes represents the dependencies between devices. Represents a node The set of adjacent nodes. Indicates the first Layer weight matrix. The weighting coefficients represent the physical information and are used to control the impact of physical coefficients on node feature updates. The node represents the result of the physical consistency regularization term. The constraints include factors such as water level in the basin and the path of flood propagation.

[0091] In this way, the features of device nodes not only obtain information from adjacent nodes, but also combine the physical constraints of the devices, such as the impact of reservoir water level and flow rate on the devices. Thus, the graph neural network fusion model (GNN) can better capture the physical relationships and scheduling needs between devices.

[0092] After multiple graph convolution operations, node features Mapped to the scheduling policy space This mapping process is accomplished through a fully connected layer (FC), as shown in the following formula (5): (5)

[0093] in, Indicates equipment The scheduling strategy indicates the operating status of the equipment (such as start-up, shutdown, opening degree, etc.). Indicates the first The node features after layer graph convolution. After the graph neural network fusion model outputs the scheduling policy for each device, the scheduling policy is sent to each corresponding device so that each device executes the corresponding scheduling policy.

[0094] It should be noted that the input data for the graph neural network fusion model includes watershed-level graphs. and device-level subgraph These graphs have captured the historical state of equipment, dependencies between equipment, and the physical health status of equipment through the features of nodes and edges. Using this data, the graph neural fusion model can process the correlation between hydrology and equipment and generate scheduling strategies. Furthermore, a physical consistency regularization term helps the graph neural fusion model better understand the physical operating patterns of equipment within the watershed, ensuring that the scheduling strategy conforms to physical constraints and further improving the accuracy of the scheduling strategy.

[0095] Steps S101 to S105, as illustrated in this embodiment, involve acquiring hydrological data, equipment status data, and equipment health data. A basin-level map is constructed based on these data. The basin-level map includes multiple equipment-level subgraphs. A physical consistency regularization term is constructed based on preset data-driven loss and preset physical consistency loss. The basin-level map, equipment-level subgraphs, and physical consistency regularization term are input into a preset graph neural network fusion model to obtain a scheduling strategy for each edge device. The scheduling strategy is then sent to each corresponding edge device. This achieves collaborative optimization scheduling across basins and among multiple devices.

[0096] Please see Figure 3 In some embodiments, after step S105, the intelligent water conservancy information management method may further include steps S301 to S304:

[0097] Step S301: Receive risk score data and optimized scheduling strategy sent by the edge device; wherein, the optimized scheduling strategy is obtained by the edge device after modifying the scheduling strategy according to the dual-sensitivity optimization model, and the risk score data is obtained by the edge-level health risk model to assess the risk of the scheduling strategy.

[0098] Step S302: Train the preset joint feedback optimization model based on the scheduling strategy, the optimized scheduling strategy, the equipment health data before the scheduling strategy is executed, the equipment health data after the scheduling strategy is executed, and the risk score data to obtain the strategy residual modeling loss and the state prediction loss.

[0099] Step S303: Update the parameters of the graph neural network fusion model according to the strategy residual modeling loss to obtain the updated graph neural network fusion model;

[0100] Step S304: Update the parameters of the edge-level health risk model and the dual-sensitivity optimization model based on the state prediction loss to obtain the updated edge-level health risk model and the updated dual-sensitivity optimization model.

[0101] In step S301 of some embodiments, after receiving the scheduling policy, the edge device performs a risk assessment on the scheduling policy to obtain risk score data. If the risk score data is too high, it indicates that the current scheduling policy will affect the healthy operation of the device. Therefore, the scheduling policy is modified according to the dual-sensitivity optimization model to obtain an optimized scheduling policy. The cloud server receives the risk score data and the optimized scheduling policy sent by the edge device.

[0102] In step S302 of some embodiments, a joint feedback optimization model is constructed based on the multi-source feedback decoupling modeling structure (MFR-Net). The joint feedback optimization model includes two major sub-models, namely the policy bias modeling branch. and post-execution effect modeling branch The policy deviation modeling branch is used to model scheduling policies. With the optimized scheduling strategy Systematic biases between them are used to optimize graph neural network fusion models. The post-execution modeling branch is used to model risk scoring data. and device health data after executing scheduling policies Inconsistency, used for optimization and dual-sensitivity optimization model The accuracy of expression. Therefore, the preset joint feedback optimization model is trained based on the scheduling strategy, the optimized scheduling strategy, the equipment health data before the scheduling strategy is executed, the equipment health data after the scheduling strategy is executed, and the risk score data, to obtain the strategy residual modeling loss and the state prediction loss. In one example, the total loss function of the joint feedback optimization model is shown in the following formula (6): (6)

[0103] in, This represents the total loss function of the joint feedback optimization model. This represents a graph neural fusion model. The generated scheduling policy, This represents the scheduling strategy optimized by the dual-sensitivity optimization model. This indicates the device health data after the scheduling policy is executed. This represents the loss from strategy residual modeling. This represents the loss from state prediction. , representing the device health data after the execution scheduling strategy is predicted by the joint feedback optimization model. This indicates the device health data before the scheduling strategy is executed. This represents risk score data.

[0104] In some embodiments, steps S303 to S304 involve modeling the loss using policy residuals. Update graph neural fusion model The parameters are used to obtain the updated graph neural fusion model. The state prediction loss is then used. Update the edge-level health risk model and the dual-sensitivity optimization model The parameters are used to obtain the updated edge-level health risk model and the updated dual-sensitivity optimized model. The updated model parameters are then distributed to the corresponding edge devices via version control.

[0105] In practical applications, datasets uploaded by edge devices are collected at fixed intervals. Update the model parameters. The dataset... All data in time series They are stored in the form of node numbers.

[0106] In this embodiment, steps S301 to S304 involve constructing a joint feedback optimization model, which is used to optimize the graph neural network fusion model based on the collected feedback data. Marginal-level health risk model and dual-sensitivity optimization model Dynamic parameter optimization is performed to improve the robustness and adaptability of the model in long-term operation.

[0107] Please refer to Figure 4 , Figure 4 This is a flowchart of the intelligent water conservancy information management method provided in the third embodiment of this application, which is applied to edge devices. Figure 4 The method may include, but is not limited to, steps S401 to S403.

[0108] Step S401: Receive the scheduling policy sent by the cloud server; wherein the scheduling policy is obtained by the cloud server based on the watershed-level graph, the device-level subgraph, and the physical consistency regularization term;

[0109] Step S402: Perform a risk assessment on the scheduling strategy based on the preset edge-level health risk model to obtain risk score data;

[0110] Step S403: If the risk score data is less than the preset first risk threshold, then the scheduling strategy is executed; if the risk score data is greater than or equal to the first risk threshold and less than the preset second risk threshold, then the amplitude limiting process is performed before executing the scheduling strategy; if the risk score data is greater than or equal to the second risk threshold, then an alarm message is sent.

[0111] In steps S401 to S402 of some embodiments, the cloud server obtains the scheduling policy based on the watershed-level graph, the device-level subgraph, and the physical consistency regularization term, and then sends it to the edge device. The edge device, relying on the edge-level health risk model deployed at each device site, achieves real-time perception of the device's current health status and a safety assessment of the scheduling policy's execution, determining whether the scheduling policy can be safely executed under the current state.

[0112] Specifically, traditional edge device anomaly detection only focuses on abnormal changes in time series (such as temperature spikes or sudden load increases). However, in hydraulic equipment, many anomaly risks originate from imbalances in the coordination between internal structures. For example, frequent alternating start-stop cycles between multiple pump groups can lead to uneven pump pressure switching and resonance risks. Inconsistent states among multiple gate actuators can cause mechanical interference during opening and closing. Therefore, a structure-aware regularization term needs to be introduced into the edge-level health risk model. This is used to measure the "policy consistency pressure" among the control units within the device. Specifically, based on the device-level subgraph and scheduling policy, a structure-aware regularization term is constructed, as shown in the following formula (7): (7)

[0113] in, This represents a structure-aware regularization term. Indicates equipment Device-level subgraph The control unit is connected to the edge. , This indicates the intensity of scheduling actions (such as opening degree and load) of the sub-control unit, from the scheduling strategy. Obtained by disassembly. The edge weight represents the coupling strength between the two units in coordinated control. The larger the value, the greater the conflict or inconsistency in the device's internal scheduling strategy, and the higher the risk.

[0114] Based on the current device health data, scheduling strategy, preset feature cross-mapping function, and structure-aware regularization term, an edge-level health risk model is constructed, as shown in the following formula (8): (8)

[0115] in, This represents risk score data, representing equipment. Execute the scheduling policy in the current state. The health risks involved range from [0,1]. Indicates the current time Device health data, such as , This represents the stress data of the dam body. This indicates the vibration data of the pump set. This indicates the temperature data of the device. This indicates the load data of the pumping station. Indicates near State trend estimation within the time window, such as . Indicates the scheduling strategy. This represents the feature cross-mapping function, used to extract the interactive influence features between scheduling policies and states. This represents a structure-aware regularization term. , This represents the model parameters after training or distillation on the edge device. This represents a structural complexity adjustment factor, which controls the degree of impact of structural risks. For the Sigmoid function, ensure the output value Mapped to [0,1].

[0116] It should be noted that a structure-aware regularization term is introduced, which is then multiplied by a coefficient. This was subsequently integrated into the main formula of the edge-level health risk model, effectively improving the "structural accuracy" of edge health assessment. The edge-level health risk model not only considers outliers in the current state of the equipment (such as high vibration and high temperature), but also models the magnitude of state disturbances caused by scheduling strategies. Structure-aware regularized design utilizes device-level subgraphs... The modeling of the coupling between control units within the equipment avoids ignoring structural risks. It boasts high edge deployability, employing local window temporal modeling and a simplified MLP inference structure, enabling both rapid operation and easy iterative optimization.

[0117] In step S403 of some embodiments, the first risk threshold can be set to 0.3, and the second risk threshold can be set to 0.7. When Then the scheduling strategy It is considered fully executable, so the scheduling policy is executed directly. This indicates a medium risk, requiring limiting measures (such as reducing the starting force) before executing the scheduling strategy. This indicates that the current state is not suitable for implementing the scheduling strategy, and it is necessary to postpone, reduce the load, or send an alarm message.

[0118] In steps S401 to S403 of this embodiment, a scheduling strategy is received from the cloud server. This scheduling strategy is derived by the cloud server based on a watershed-level map, a device-level sub-map, and a physical consistency regularization term. A risk assessment is performed on the scheduling strategy according to a preset edge-level health risk model, resulting in risk score data. If the risk score data is less than a preset first risk threshold, the scheduling strategy is executed. If the risk score data is greater than or equal to the first risk threshold but less than a preset second risk threshold, a limiting process is performed before executing the scheduling strategy. If the risk score data is greater than or equal to the second risk threshold, an alarm message is sent. From the generation of the scheduling strategy to its execution by the edge device, a physically perceptible, status-determinable, and risk-controllable security collaboration mechanism is formed, laying a crucial foundation for the long-term healthy operation of water conservancy equipment and the stability of the scheduling system.

[0119] Please see Figure 5 In some embodiments, if the risk score data is greater than or equal to the second risk threshold, the intelligent water conservancy information management method may also include, but is not limited to, steps S501 to S502:

[0120] Step S501: Construct a dual-sensitivity optimization model based on the scheduling strategy, risk score data, current device health data, and structure-aware regularization term;

[0121] Step S502: The scheduling strategy is modified according to the dual-sensitivity optimization model to obtain the optimized scheduling strategy.

[0122] In step S501 of some embodiments, a dual-sensitivity optimization model for edge deployment is constructed. To achieve this based on risk scoring data Current device health data and structure-aware regularization For scheduling strategies Perform parameterized optimization and output the optimized scheduling strategy. The dual-sensitivity optimization model will be deployed long-term at the edge, possessing the characteristics of lightweight structure, online inference capability, and local fine-tuning capability. The core of the dual-sensitivity optimization model is a two-branch shallow network that handles the "risk adjustment path" and the "structure adjustment path" respectively, and merges the output policy residuals to achieve policy correction. The structure of the dual-sensitivity optimization model is shown in the following formulas (9) and (10): (9) (10)

[0123] in, This indicates the optimized scheduling strategy. Indicates the scheduling strategy. This represents the residual from strategy optimization, which is the sum of the outputs of the two sub-branches. This represents risk score data. This indicates the device's health data at the current moment. This represents a structure-aware regularization term. Indicates the risk adjustment branch, input and Learn about the impact of risks on strategy residuals under the current health status. Indicates a structural adjustment branch, input and The optimization direction of strategy based on the collaborative risk of learning structure. , This represents the branch fusion weight matrix, which is learned during training.

[0124] It should be noted that the risk adjustment branch within the model... The focus is on whether the magnitude of the prediction scheduling strategy is too extreme and whether the scheduling is too frequent, based on the input. and Establish a perception-response relationship. For example, if (Medium-high risk), while load (Approaching the limit), temperature rise Then the risk adjustment branch The output strategy residual may be [-0.2, -0.15, 0.0] (downward adjustment sub-instruction). Simultaneously, the structure adjustment branch... according to Insufficient coordination between pump units was detected, and the output residual [0.0, +0.1, -0.1] (balancing the operating amplitude between pump units) was calculated. The two values ​​are then added together to obtain the final value. The output is used to correct the scheduling strategy.

[0125] In steps S501 to S502 of this embodiment, when the risk score data is greater than or equal to the second risk threshold and the current scheduling strategy cannot be executed, the scheduling strategy is modified using a dual-sensitivity optimization model to obtain an optimized scheduling strategy. This enables the edge device to execute the optimized scheduling strategy, ensuring the long-term healthy operation of the device.

[0126] Please see Figure 6 This application also provides an intelligent water conservancy information management platform, applied to a cloud server, which can realize the above-mentioned intelligent water conservancy information management method applied to a cloud server. The system includes:

[0127] The acquisition module 601 is used to acquire hydrological data, equipment status data, and equipment health data.

[0128] The first construction module 602 is used to construct a watershed-level map based on hydrological data, equipment status data, and equipment health data; wherein the watershed-level map includes multiple equipment-level sub-maps;

[0129] The second construction module 603 is used to construct a physical consistency regularization term based on a preset data-driven loss and a preset physical consistency loss.

[0130] The input module 604 is used to input the watershed-level graph, the device-level subgraph, and the physical consistency regularization term into the preset graph neural network fusion model to obtain the scheduling strategy for each edge device; wherein, the scheduling strategy is used to indicate the operating status of the edge device;

[0131] The sending module 605 is used to send the scheduling policy to each corresponding edge device.

[0132] The specific implementation method of this intelligent water conservancy information management platform is basically the same as the specific implementation method of the intelligent water conservancy information management method applied to cloud servers mentioned above, and will not be repeated here.

[0133] Please see Figure 7 This application also provides an intelligent water conservancy information management platform for edge devices, which can realize the above-mentioned intelligent water conservancy information management method for edge devices. The system includes:

[0134] The receiving module 701 is used to receive the scheduling policy sent by the cloud server; wherein the scheduling policy is obtained by the cloud server based on the watershed level graph, the device level subgraph and the physical consistency regularization term;

[0135] The evaluation module 702 is used to perform risk assessment on the scheduling strategy based on a preset edge-level health risk model and obtain risk score data.

[0136] The execution module 703 is used to execute a scheduling strategy if the risk score data is less than a preset first risk threshold, perform a limit-down process before executing the scheduling strategy if the risk score data is greater than or equal to the first risk threshold and less than a preset second risk threshold, and send an alarm message if the risk score data is greater than or equal to the second risk threshold.

[0137] The specific implementation method of this intelligent water conservancy information management platform is basically the same as the specific implementation method of the intelligent water conservancy information management method applied to edge devices described above, and will not be repeated here.

[0138] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned intelligent water conservancy information management method applied to a central server and the intelligent water conservancy information management method applied to an edge device. This electronic device can be any intelligent terminal, including tablet computers, in-vehicle computers, etc.

[0139] Please see Figure 8 , Figure 8 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes:

[0140] The processor 801 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application.

[0141] The memory 802 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 802 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 802 and called and executed by the processor 801. The methods described in this application for intelligent water conservancy information management applied to a central server and intelligent water conservancy information management applied to edge devices are described below.

[0142] The 803 input / output interface is used to implement information input and output.

[0143] The communication interface 804 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0144] Bus 805 transmits information between various components of the device (e.g., processor 801, memory 802, input / output interface 803, and communication interface 804);

[0145] The processor 801, memory 802, input / output interface 803, and communication interface 804 are connected to each other within the device via bus 805.

[0146] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned intelligent water conservancy information management method applied to a central server and the intelligent water conservancy information management method applied to an edge device.

[0147] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0148] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0149] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0150] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0151] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0152] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0153] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0154] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of the system or units may be electrical, mechanical, or other forms.

[0155] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0156] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0157] If the integrated unit is implemented as 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 this application, in essence, 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. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0158] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. An intelligent water information management method, characterized in that, Applied to cloud servers, the method includes: Acquire hydrological data, equipment status data, and equipment health data; A watershed-level map is constructed based on the hydrological data, equipment status data, and equipment health data; wherein, the watershed-level map includes multiple equipment-level subgraphs; constructing the watershed-level map based on the hydrological data, equipment status data, and equipment health data includes: mapping the hydrological data, equipment status data, and equipment health data to the watershed-level map, wherein the nodes of the watershed-level map are various equipment in the watershed, and the edges are the dependencies between equipment; each equipment node in the watershed-level map is a device-level subgraph, and the device-level subgraph contains multiple units within the equipment and their control relationships; Based on the preset data-driven loss and the preset physical consistency loss, a physical consistency regularization term is constructed; The watershed-level graph, the device-level subgraph, and the physical consistency regularization term are input into a preset graph neural network fusion model to obtain a scheduling strategy for each edge device; wherein, the scheduling strategy is used to indicate the operating status of the edge device; The scheduling policy is sent to each of the corresponding edge devices.

2. The method of claim 1, wherein, After sending the scheduling policy to each of the corresponding edge devices, the method further includes: The system receives risk score data and the optimized scheduling strategy sent by the edge device; wherein the optimized scheduling strategy is obtained by the edge device after modifying the scheduling strategy according to the dual-sensitivity optimization model, and the risk score data is obtained by the edge-level health risk model to assess the risk of the scheduling strategy. The preset joint feedback optimization model is trained based on the scheduling strategy, the optimized scheduling strategy, the equipment health data before the scheduling strategy is executed, the equipment health data after the scheduling strategy is executed, and the risk score data to obtain the strategy residual modeling loss and the state prediction loss. The graph neural network fusion model is updated with parameters based on the strategy residual modeling loss to obtain the updated graph neural network fusion model. The parameters of the edge-level health risk model and the dual-sensitivity optimization model are updated based on the state prediction loss to obtain the updated edge-level health risk model and the updated dual-sensitivity optimization model.

3. The method of claim 1, wherein, The step of constructing a physical consistency regularization term based on a preset data-driven loss and a preset physical consistency loss includes: Obtain actual observation data; The data-driven loss is constructed based on the error between the scheduling strategy and the actual observed data. To ensure that the scheduling strategy conforms to the physical constraints of the edge device, the physical consistency loss is constructed. The physical consistency regularization term is constructed based on the data-driven loss and the physical consistency loss.

4. An intelligent water information management method, characterized in that, Applied to edge devices, the method includes: The system receives a scheduling strategy sent by a cloud server. This scheduling strategy is obtained by the cloud server based on a watershed-level graph, a device-level subgraph, and a physical consistency regularization term. The cloud server maps hydrological data, device status data, and device health data to the watershed-level graph. The nodes of the watershed-level graph represent various devices within the watershed, and the edges represent dependencies between devices. Each device node in the watershed-level graph is a device-level subgraph, which contains multiple units within the device and their control relationships. The scheduling strategy is risk-assessed based on a preset edge-level health risk model to obtain risk score data. If the risk score data is less than a preset first risk threshold, the scheduling strategy is executed. If the risk score data is greater than or equal to the first risk threshold and less than a preset second risk threshold, the amplitude is limited before the scheduling strategy is executed. If the risk score data is greater than or equal to the second risk threshold, an alarm message is sent.

5. The method of claim 4, wherein, The method further includes: constructing the edge-level health risk model, specifically including: Based on the device-level subgraph and the scheduling strategy, construct a structure-aware regularization term; The edge-level health risk model is constructed based on the current device health data, the scheduling strategy, the preset feature cross-mapping function, and the structure-aware regularization term.

6. The method of claim 5, wherein, If the risk score data is greater than or equal to the second risk threshold, the method further includes: A dual-sensitivity optimization model is constructed based on the scheduling strategy, the risk scoring data, the current device health data, and the structure-aware regularization term. The scheduling strategy is modified according to the dual-sensitivity optimization model to obtain the optimized scheduling strategy.

7. An intelligent water information management platform, characterized in that, The management platform, applied to cloud servers, includes: The acquisition module is used to acquire hydrological data, equipment status data, and equipment health data. The first construction module is used to construct a watershed-level map based on the hydrological data, the equipment status data, and the equipment health data; wherein the watershed-level map includes multiple equipment-level subgraphs; constructing the watershed-level map based on the hydrological data, the equipment status data, and the equipment health data includes: mapping the hydrological data, the equipment status data, and the equipment health data to the watershed-level map, wherein the nodes of the watershed-level map are various equipment in the watershed, and the edges are the dependencies between equipment; each equipment node in the watershed-level map is a device-level subgraph, and the device-level subgraph contains multiple units within the equipment and their control relationships; The second construction module is used to construct a physical consistency regularization term based on the preset data-driven loss and the preset physical consistency loss. The input module is used to input the watershed-level graph, the device-level subgraph, and the physical consistency regularization term into a preset graph neural network fusion model to obtain the scheduling strategy for each edge device; wherein, the scheduling strategy is used to indicate the operating status of the edge device; The sending module is used to send the scheduling policy to each of the corresponding edge devices.

8. An intelligent water information management platform, characterized in that, The management platform, applied to edge devices, includes: A receiving module is used to receive a scheduling strategy sent by a cloud server. The scheduling strategy is obtained by the cloud server based on a watershed-level graph, a device-level subgraph, and a physical consistency regularization term. The cloud server maps hydrological data, device status data, and device health data to the watershed-level graph. The nodes of the watershed-level graph represent various devices in the watershed, and the edges represent dependencies between devices. Each device node in the watershed-level graph is a device-level subgraph, which contains multiple units within the device and their control relationships. The evaluation module is used to evaluate the scheduling strategy based on a preset edge-level health risk model and obtain risk score data. The execution module is configured to execute the scheduling strategy if the risk score data is less than a preset first risk threshold, perform amplitude limiting processing before executing the scheduling strategy if the risk score data is greater than or equal to the first risk threshold and less than a preset second risk threshold, and send an alarm message if the risk score data is greater than or equal to the second risk threshold.

9. An electronic device, comprising: The electronic device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the intelligent water conservancy information management method according to any one of claims 1 to 3 and the intelligent water conservancy information management method according to any one of claims 4 to 6.

10. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 9. When the computer program is executed by the processor, it implements the intelligent water conservancy information management method according to any one of claims 1 to 3 and the intelligent water conservancy information management method according to any one of claims 4 to 6.

Images