Multi-gate cooperative control system and control method thereof

Abstract

The invention discloses a multi-gate cooperative control system and a control method thereof, and relates to the field of water conservancy information automation and intelligent control. Comprising the following steps: collecting multi-gate hydrological state data in real time, and carrying out numerical value validity detection and logic judgment; preprocessing the multi-gate hydrological state data after logical judgment to obtain structured multi-gate hydrological state data; based on the structured multi-gate hydrological state data, predicting the future hydrological state of each gate node to obtain a structured hydrological state prediction result; generating three scoring indexes based on the structured hydrological state prediction result; weighting the three scoring indexes to obtain a comprehensive score; and based on the comprehensive score, a multi-gate collaborative optimal opening strategy is generated, gate actions are controlled, multi-gate hydrological state data are collected in real time, and a closed loop is achieved. The problems that a traditional gate control system lacks a cooperative scheduling mechanism, is insufficient in predictive capacity, lags in control response and is rigid in regulation and control strategy are solved.

Description

Technical Field

[0001] The present invention relates to the field of water conservancy information automation and intelligent control, and particularly relates to a multi-gate collaborative control system and a control method thereof. Background Art

[0002] With the continuous improvement of the level of water conservancy informatization and automation, gate control, as a core link in basin dispatching, water resources management, flood control and drainage, etc., its intelligent and collaborative regulation ability has become an important direction for the current system upgrade. Most traditional gate control systems adopt a local independent control architecture, and the control logic is mainly based on threshold setting and manual experience. However, the existing gate control systems still have the following problems: (1) There is a lack of a linkage coordination mechanism between distributed systems; (2) The system response is lagged and the emergency response ability is weak; (3) The scheduling algorithm is single and lacks an intelligent prediction mechanism; (4) The control target is single and lacks multi-target scheduling optimization; (5) There is no state feedback mechanism and the control closed-loop ability is insufficient.

[0003] In summary, there is an urgent need for a multi-gate collaborative control system with the abilities of intelligent prediction, flexible scheduling, linkage coordination and state feedback to realize the intelligent modeling, prediction-driven, closed-loop scheduling and linkage execution of the multi-gate collaborative control system under complex hydrological environments. Summary of the Invention

[0004] The present invention provides a multi-gate collaborative control system and a control method thereof to solve the technical problems of the traditional gate control system lacking a collaborative scheduling mechanism, insufficient prediction ability, lagging control response and rigid regulation strategy.

[0005] The multi-gate collaborative control system and the control method thereof of the present invention specifically include the following technical solutions: A multi-gate collaborative control method includes the following steps: S1. Real-time collect the hydrological state data of multiple gates, and perform numerical validity detection and logical judgment on the hydrological state data of multiple gates; preprocess the hydrological state data of multiple gates after logical judgment to obtain structured hydrological state data of multiple gates; S2. Based on the structured hydrological state data of multiple gates, through a graph neural network algorithm optimized by a graph attention mechanism, predict the future hydrological states of each gate node to obtain a structured hydrological state prediction result; S3. Based on the structured hydrological state prediction result, construct a state scoring model to generate a water level target score, a fluctuation stability score and an action energy consumption score as scoring indicators; weight the water level target score, the fluctuation stability score and the action energy consumption score to obtain a comprehensive score; based on the comprehensive score, generate an optimal opening strategy for multi-gate collaboration, control the gate action, and real-time collect the hydrological state data of multiple gates to achieve a closed loop.

[0006] Preferably, the S1 specifically includes: The multi-gate hydrological state data includes hydrological data and equipment state data, where the hydrological data includes the upstream water level, downstream water level, and flow velocity; the equipment state data includes the motor operating current and the gate opening.

[0007] Preferably, the S1 specifically includes: Perform numerical validity detection on the multi-gate hydrological state data to check whether the multi-gate hydrological state data is within the set range; based on the multi-gate hydrological state data after numerical validity detection, perform logical judgment. When the logic is normal, the multi-gate collaborative control system will maintain the current stable operating state; when the logic is abnormal, the local emergency control logic will be triggered to execute a fast control response.

[0008] Preferably, the S2 specifically includes: In the graph neural network algorithm optimized based on the graph attention mechanism, construct the multi-gate collaborative control system as a dynamic directed graph, split the structured multi-gate hydrological state data according to the gate nodes, and standardize it to generate the feature vectors of the gate nodes.

[0009] Preferably, the S2 specifically includes: Based on the feature vectors of the gate nodes, propose an edge weight attention mechanism on the basis of the graph neural network algorithm, define the edge attention weights, and adjust the influence intensity between each gate node according to the real-time hydrological state.

[0010] Preferably, the S2 specifically includes: On the basis of the edge weight attention mechanism, propose a time series attention mechanism to obtain the predicted features of the gate nodes at the next moment; input the predicted features into a long short-term memory neural network for feature recognition, and output the predicted values of the hydrological variables of the target gate nodes as the structured hydrological state prediction results.

[0011] Preferably, the S3 specifically includes: Generate a water level target score based on the relative error between the upstream water level of the gate node and the set target water level value; generate a fluctuation smoothness score based on the difference in the upstream water levels of the gate node at different times, which is used to measure the smoothness of the water level prediction sequence; generate an action energy consumption score based on the difference in the gate openings of the gate node at different times.

[0012] Preferably, the S3 specifically includes: Construct an alternative opening strategy combination, calculate the comprehensive score of each opening strategy combination within the future prediction time window, and select the opening strategy combination with the highest comprehensive score as the optimal opening strategy for multi-gate collaboration in the current cycle.

[0013] A multi-gate collaborative control system includes the following parts: Gate execution and perception module, edge control and processing module, data processing module, graph modeling and prediction module, policy calculation module; The gate execution and perception module is used to control the gate actions and collect multi-gate hydrological state data in real time, and transfer the multi-gate hydrological state data to the edge control and processing module; and receive the multi-gate collaborative optimal opening strategy issued by the edge control and processing module to drive the gate actuator in the gate execution and perception module to control the gate actions; The edge control and processing module is used to perform emergency control of the gates, and at the same time complete the validity detection and logical judgment of the multi-gate hydrological state data to obtain the processed multi-gate hydrological state data, and upload the processed multi-gate hydrological state data to the data processing module; at the same time, receive the multi-gate collaborative optimal opening strategy sent by the policy calculation module and send the multi-gate collaborative optimal opening strategy to the gate execution and perception module; The data processing module is used to perform structured processing on the multi-gate hydrological state data processed by the edge control and processing module to form structured multi-gate hydrological state data, and transfer the structured multi-gate hydrological state data to the graph modeling and prediction module; The graph modeling and prediction module obtains the structured hydrological state prediction result through a graph neural network algorithm optimized based on the graph attention mechanism, and outputs the structured hydrological state prediction result to the policy calculation module; The policy calculation module is used to generate a multi-gate collaborative optimal opening strategy through a state scoring model according to the structured hydrological state prediction result of the graph modeling and prediction module, and send the multi-gate collaborative optimal opening strategy to the edge control and processing module.

[0014] The beneficial effects of the technical solution of the present invention are: 1. Deploy the preliminary logical judgment and emergency control functions in the edge control and processing module, which can independently complete the closed-loop control in case of network interruption or sudden working conditions, avoid the scheduling risks caused by cloud delay or interruption, and greatly improve the stability and safety of the multi-gate collaborative control system operation.

[0015] 2. Construct a dynamic directed graph structure with gates as nodes and hydraulic regulation effects as edges, and introduce the edge weight attention mechanism and time series attention mechanism, and propose a graph neural network algorithm optimized based on the graph attention mechanism, which significantly enhances the trend modeling and prediction accuracy of the multi-gate collaborative control system hydrological state.

[0016] 3. Through the state scoring model, including water level target scoring, fluctuation stability scoring and action energy consumption scoring, the scheduling strategy screening and optimization under prediction are realized, the transparency, debuggability and target adaptation ability of the scheduling strategy are improved, and the multi-objective fusion control such as flood control, ecology and energy saving is realized. Description of the Drawings

[0017] Figure 1 Structural diagram of a multi-gate collaborative control system according to the present invention; Figure 2 Flowchart of a multi-gate collaborative control method according to the present invention. Detailed implementation manners

[0018] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0020] The following specifically describes in conjunction with the accompanying drawings the specific solutions of a multi-gate collaborative control system and its control method provided by the present invention.

[0021] Refer to the attached Figure 1 , which shows a structural diagram of a multi-gate collaborative control system provided by an embodiment of the present invention. The system includes the following parts: Gate execution and sensing module, edge control processing module, data processing module, graph modeling and prediction module, strategy calculation module; The gate execution and sensing module is used to control the gate action and collect multi-gate hydrological state data in real time, and transfer the collected multi-gate hydrological state data to the edge control processing module; and receive the multi-gate collaborative optimal opening strategy sent by the edge control processing module to drive the gate execution mechanism in the gate execution and sensing module to control the gate action; The edge control processing module is used to perform emergency control of the gate in case of network interruption or emergency, and at the same time complete the validity detection and logical judgment of the multi-gate hydrological state data to obtain the processed multi-gate hydrological state data, and upload the processed multi-gate hydrological state data to the data processing module; at the same time, receive the multi-gate collaborative optimal opening strategy sent by the strategy calculation module and transfer the multi-gate collaborative optimal opening strategy to the gate execution and sensing module; The data processing module is used to perform structured processing on the multi-gate hydrological state data processed by the edge control processing module to form structured multi-gate hydrological state data, and transfer the structured multi-gate hydrological state data to the graph modeling and prediction module; The graph modeling prediction module obtains the structured hydrological state prediction result through the graph neural network algorithm optimized based on the graph attention mechanism, and outputs the structured hydrological state prediction result to the policy calculation module; The policy calculation module is used to generate the multi-gate collaborative optimal opening policy through the state scoring model according to the structured hydrological state prediction result of the graph modeling prediction module, and send the multi-gate collaborative optimal opening policy to the edge control processing module.

[0022] Refer to the appendix Figure 2 , which shows a flowchart of a multi-gate collaborative control method provided by an embodiment of the present invention. The method includes the following steps: S1. Real-time collect the multi-gate hydrological state data, and perform numerical validity detection and logical judgment on the multi-gate hydrological state data; preprocess the multi-gate hydrological state data after logical judgment to obtain the structured multi-gate hydrological state data.

[0023] S101. Real-time collect the multi-gate hydrological state data through devices such as sensors deployed at each gate. The devices such as sensors include: ultrasonic water level sensors, current sensors, electromagnetic flow meters, and position encoders; the multi-gate hydrological state data includes hydrological data and device status data, where the hydrological data includes the upstream water level and the downstream water level, unit: meter (m), and the flow rate, unit: meter per second (m / s); the device status data includes the motor working current, unit: ampere (A), and the gate opening, unit: percentage (%).

[0024] After the gate execution and perception module uploads the collected multi-gate hydrological state data to the edge control and processing module, the edge control and processing module first conducts a numerical validity check on the multi-gate hydrological state data: it checks whether the multi-gate hydrological state data is within the set range, and the set range is determined based on the expert experience method. For example, the validity range of the upstream water level designed for the multi-gate collaborative control system is from 1.0 meter to 9.5 meters; the validity range of the flow velocity is from 0.1 meter per second to 4.8 meters per second; and the moving average and range judgment method is used to remove noise. Then, based on the multi-gate hydrological state data after the numerical validity check, a logical judgment is made: if the logic is normal, the multi-gate collaborative control system will maintain the current stable operation state, and the edge control and processing module will only upload the multi-gate hydrological state data after the numerical validity check to the data processing module at a certain period. The setting of the period depends on the frequency of hydrological state changes, the real-time requirements of the multi-gate collaborative control system, and the bandwidth of the communication system, and can be set to reasonable time intervals such as 5 minutes, 10 minutes, or 30 minutes, etc.; if the logic is abnormal, the local emergency control logic in the edge control and processing module will be triggered to execute a fast control response. The logical judgment is based on the expert experience method. For example, experts may set a water level-flow velocity relationship model according to experience. If the flow velocity and water level changes do not match, it is considered that the logic is abnormal.

[0025] Finally, the multi-gate hydrological state data processed by the edge control and processing module is uploaded to the data processing module.

[0026] The local emergency control logic under abnormal logic specifically includes but is not limited to: ① If the water level difference between the upstream and downstream of a certain gate exceeds 0.5 meter within 5 minutes, the multi-gate collaborative control system will automatically execute multiple response strategies such as status marking, opening the pre-discharge mode, triggering a warning, and restricting the subsequent opening adjustment rhythm; ② If the motor runs continuously for an excessive time or the gate gets stuck, enter the "equipment protection mode" and send a text message / light and sound alarm prompt; ③ If the upstream water level is higher than the set flood discharge warning line, such as the set flood discharge warning line is 5.20 meters, start the "safety pre-discharge strategy" and automatically adjust the gate opening to the preset safe opening (such as 60%) to reduce the upstream water level.

[0027] S102. After the data processing module receives the multi-gate hydrological state data processed by the edge control and processing module, it performs preprocessing operations on the multi-gate hydrological state data processed by the edge control and processing module to form structured multi-gate hydrological state data; the preprocessing operations include time alignment, cleaning, structuring, standardization, etc., which are all well-known technical means to those skilled in the art; finally, the structured multi-gate hydrological state data is sent to the graph modeling and prediction module.

[0028] S2. Based on the structured multi-gate hydrological state data, through the graph neural network algorithm optimized by the graph attention mechanism, predict the future hydrological state of each gate node to obtain the structured hydrological state prediction result.

[0029] S201. Based on the structured multi-gate hydrological state data, propose a graph neural network algorithm optimized by the graph attention mechanism, and construct the entire multi-gate collaborative control system as a dynamic directed graph , where represents the set of gate nodes, represents the th gate node, represents the number of gate nodes; the edge set consists of node pairs with actual hydraulic regulation influence relationships between gates. Each edge represents the actual influence direction of gate node on gate node in terms of hydraulic regulation. Gate node is the neighbor gate node of gate node .

[0030] Split the structured multi-gate hydrological state data according to the gate nodes, and standardize it to generate the feature vector of gate node . The feature vector includes: the upstream water level at time ; the downstream water level at time ; the gate opening at time ; the flow velocity at time ; the motor working current at time .

[0031] S202. The influence between gates is not static, but is affected by the current water level difference, water flow change, gate position, etc. Therefore, an edge weight attention mechanism improvement method is proposed based on the graph neural network algorithm, and the edge attention weight is defined as: , where the edge attention weight represents the influence weight of gate node on gate node at time ; is a normalization function used to normalize the attention scores of neighbor gate nodes into a weight distribution; represents the attention score weight vector, which is obtained through experiments and is used for the concatenated feature combination Perform a linear scoring. The higher the score, the more attention the gate node pays to the gate node ; represents the transpose of the attention scoring weight vector ; represents an activation function, such as ReLU or LeakyReLU, which is used to perform a non-linear mapping on the result of the linear combination to enhance the expressive ability of the edge weight attention mechanism; represents the edge feature mapping weight matrix, obtained through experiments, which is used to perform a linear transformation on the concatenated node and edge features; represents the gate node at time; represents the gate node at time; , represents the hydraulic transmission intensity at time; represents the vector concatenation operation.

[0032] By introducing the edge weight attention mechanism, the graph neural network algorithm optimized based on the graph attention mechanism can adaptively adjust the influence intensity between each gate node according to the real-time hydrological state, so as to more accurately describe the hydraulic conduction relationship between the gates and improve the prediction accuracy and generalization ability under complex hydrological conditions (such as sudden floods and asynchronous regulation).

[0033] S203. Since the hydrological evolution of the multi-gate collaborative control system also has time correlation and time series cumulative effect, a time series attention mechanism is further proposed on the basis of the edge weight attention mechanism.

[0034] Define the predicted feature of the gate node at the next moment, that is, time, as: , where represents the predicted feature representation of the gate node at time; represents the set of neighbor gate nodes of the gate node ; represents an activation function, such as ReLU or LeakyReLU, which is responsible for performing non-linear processing on the aggregation result to enhance the node feature expression ability; the edge attention weight represents that at time, the gate node to the gate node influence weight; represents the feature mapping weight matrix for temporal state evolution, which is used to project the feature vector at moment of the neighbor gate node onto a mapping space, obtained from experiments; represents the feature vector of the gate node at moment.

[0035] By introducing the time series attention mechanism, it is possible to capture the leading role of the historical multi-gate hydrological state data at critical moments in the evolution trend of the current target gate node, improving the prediction accuracy and robustness of the multi-gate collaborative control system under complex temporal backgrounds. The historical multi-gate hydrological state data is taken from the database.

[0036] S204. Input the predicted features into the existing long short-term memory neural network for feature recognition. The long short-term memory neural network finally outputs the predicted values of the hydrological variables of the target gate node. The predicted values of the hydrological variables of the target gate node include the upstream water level at moment, downstream water level gate opening and flow velocity as the structured hydrological state prediction result for output. The existing long short-term memory neural network technology is relatively mature, and the existing technology can be directly adopted.

[0037] S3. Based on the structured hydrological state prediction result, construct a state scoring model to generate a water level target score, a fluctuation stability score, and an action energy consumption score as scoring indicators; weight the water level target score, the fluctuation stability score, and the action energy consumption score to obtain a comprehensive score; based on the comprehensive score, generate an optimal multi-gate collaborative opening strategy to control the gate action, and collect the multi-gate hydrological state data in real time to achieve a closed loop.

[0038] S301. Based on the structured hydrological state prediction result, the strategy calculation module constructs a state scoring model for the gate node to generate three scoring indicators, including: water level target score, fluctuation stability score, and action energy consumption score. The specific formulas are as follows: , , , where represents the water level target score, which is used to calculate the gate node at The upstream water level at a certain moment and the target water level value The relative error between them. The smaller the error, the higher the water level target score, and the score approaches 1; Γ is the number of predicted time steps; is the gate node at the upstream water level at a certain moment; is the target water level value set according to expert experience; represents the fluctuation smoothness score, which is used to measure the smoothness of the water level prediction sequence. If the water level changes slowly in the future, the higher the fluctuation smoothness score, and the score approaches 1; is the gate node at the upstream water level at a certain moment; is the gate node at the upstream water level at a certain moment; represents the action energy consumption score, which is used to calculate the gate opening of the gate node at the gate opening at a certain moment and the gate opening at a certain moment The smaller the difference, the higher the action energy consumption score, and the score approaches 1; represents the maximum allowable change range of the gate opening, which is used for normalization. It is set to 100, indicating that the adjustment amplitude is calculated as a percentage.

[0039] S302. Perform weighted calculation on the water level target score, fluctuation smoothness score, and action energy consumption score to obtain the comprehensive score : , where , , are the weight coefficients, which are set according to the expert experience method. Further construct several alternative opening strategy combinations, such as: the flood control priority strategy combination , the energy saving priority strategy combination .

[0040] Calculate the comprehensive score of each opening strategy combination within the future prediction time window, and select the opening strategy combination with the highest comprehensive score as the multi-gate collaborative optimal opening strategy for the current period.

[0041] By introducing the state scoring model, the comprehensive scoring and strategy screening of the future behavior of multiple gates are realized, taking into account water level control, stability and operation cost, and the multi-gate collaborative scheduling execution process with interpretability, flexibility and engineering feasibility is realized.

[0042] In S303, the policy calculation module sends the multi-gate collaborative optimal opening policy to the edge control processing module. After receiving the multi-gate collaborative optimal opening policy, the edge control processing module drives the gate actuators in the sensing module to control the gate actions according to the specified opening, and real-time collects the multi-gate hydrological state data and uploads it to the edge control processing module to achieve a closed loop.

[0043] In summary, a multi-gate collaborative control system and its control method are completed.

[0044] The sequence of the invention embodiments is only for description and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0045] Each embodiment in this specification is described in a progressive manner. For the same or similar parts between the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments.

[0046] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A multi-gate collaborative control method, characterized in that, It includes the following steps: S1. Collect multi-gate hydrological state data in real time, and perform numerical validity detection and logical judgment on the multi-gate hydrological state data; Preprocess the multi-gate hydrological state data after logical judgment to obtain structured multi-gate hydrological state data; S2. Based on the structured multi-gate hydrological state data, through the graph neural network algorithm optimized by the graph attention mechanism, predict the future hydrological state of each gate node to obtain a structured hydrological state prediction result; S3. Based on the structured hydrological state prediction result, construct a state scoring model to generate a water level target score, a fluctuation stability score, and an action energy consumption score as scoring indicators; weight the water level target score, the fluctuation stability score, and the action energy consumption score to obtain a comprehensive score; based on the comprehensive score, generate an optimal multi-gate collaborative opening strategy, control the gate action, and collect multi-gate hydrological state data in real time to achieve a closed loop.

2. The multi-gate collaborative control method according to claim 1, characterized in that The S1 specifically includes: The multi-gate hydrological state data includes hydrological data and equipment status data, where the hydrological data includes the upstream water level, the downstream water level, and the flow velocity; the equipment status data includes the motor working current and the gate opening.

3. The multi-gate collaborative control method according to claim 2, wherein The S1 specifically includes: Perform numerical validity detection on the multi-gate hydrological state data to check whether the multi-gate hydrological state data is within the set range; perform logical judgment based on the multi-gate hydrological state data after numerical validity detection. When the logic is normal, the multi-gate collaborative control system will maintain the current stable operation state; when the logic is abnormal, the local emergency control logic will be triggered to execute a fast control response.

4. The multi-gate collaborative control method according to claim 1, wherein The S2 specifically includes: In the graph neural network algorithm optimized by the graph attention mechanism, construct the multi-gate collaborative control system as a dynamic directed graph, split the structured multi-gate hydrological state data according to the gate nodes, and standardize to generate the feature vectors of the gate nodes.

5. The multi-gate collaborative control method according to claim 4, characterized in that, The S2 specifically includes: Based on the feature vectors of the gate nodes, propose an edge weight attention mechanism on the basis of the graph neural network algorithm, define the edge attention weights, and adjust the influence intensity between each gate node according to the real-time hydrological state.

6. The multi-gate collaborative control method according to claim 5, wherein The S2 specifically includes: On the basis of the edge weight attention mechanism, propose a time series attention mechanism to obtain the predicted features of the gate nodes at the next moment; input the predicted features into a long short-term memory neural network for feature recognition, and output the predicted values of the hydrological variables of the target gate nodes as the structured hydrological state prediction result.

7. The multi-gate collaborative control method according to claim 1, wherein The S3 specifically includes: Generate a water level target score based on the relative error between the upstream water level of the gate node and the set target water level value; generate a fluctuation stability score based on the difference in the upstream water level of the gate node at different times to measure the stability of the water level prediction sequence; generate an action energy consumption score based on the difference in the gate opening of the gate node at different times.

8. The multi-gate collaborative control method according to claim 1, wherein The S3 specifically includes: Construct an alternative opening strategy combination, calculate the comprehensive score of each opening strategy combination within the future prediction time window, and select the opening strategy combination with the highest comprehensive score as the optimal multi-gate collaborative opening strategy for the current cycle.

9. A multi-gate collaborative control system, applied to the multi-gate collaborative control method described in claim 1, characterized in that, It includes the following parts: Gate execution and perception module, edge control and processing module, data processing module, graph modeling and prediction module, policy calculation module; The gate execution and perception module is used to control the gate actions and collect multi-gate hydrological state data in real time, and transfer the multi-gate hydrological state data to the edge control and processing module; and receive the multi-gate collaborative optimal opening strategy sent down by the edge control and processing module to drive the gate execution mechanism in the gate execution and perception module to control the gate actions; The edge control and processing module is used to perform emergency control of the gates, and at the same time complete the validity detection and logical judgment of the multi-gate hydrological state data to obtain the processed multi-gate hydrological state data, and upload the processed multi-gate hydrological state data to the data processing module; at the same time, receive the multi-gate collaborative optimal opening strategy sent down by the policy calculation module and send the multi-gate collaborative optimal opening strategy to the gate execution and perception module; The data processing module is used to perform structured processing on the multi-gate hydrological state data processed by the edge control and processing module to form structured multi-gate hydrological state data, and transfer the structured multi-gate hydrological state data to the graph modeling and prediction module; The graph modeling and prediction module obtains the structured hydrological state prediction result through the graph neural network algorithm optimized based on the graph attention mechanism, and outputs the structured hydrological state prediction result to the policy calculation module; The policy calculation module is used to generate the multi-gate collaborative optimal opening strategy through the state scoring model according to the structured hydrological state prediction result of the graph modeling and prediction module, and send the multi-gate collaborative optimal opening strategy to the edge control and processing module.

Images