Multi-water-source joint scheduling method and system

Abstract

The invention discloses a multi-water-source joint scheduling method and system, and relates to the technical field of water resource scheduling, and the method comprises the steps: setting a flood control water level target, and calculating the real-time water level data of each reservoir and the water level deviation value of the flood control water level target; predicting the water discharge amount of each reservoir by using a reinforcement learning algorithm, and adjusting the predicted water discharge amount based on the real-time water level change trend; a cooperative control mechanism between the reservoirs is constructed, gate flow parameters of all the reservoirs are calculated through a time sequence algorithm, and a reservoir gate control model is constructed based on the gate flow parameters; generating a gate control instruction in combination with the reservoir gate control model; and inputting the generated gate control instruction into a gate controller, and in combination with a preset scheduling rule, executing water source scheduling between reservoirs. The reinforcement learning algorithm is used for predicting the water discharge amount, dynamic adjustment is conducted according to the real-time water level change trend, unnecessary water resource waste is avoided, and the risk of isolated operation is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of water resource scheduling, and in particular to a multi-water source joint scheduling method and system. Background Art

[0002] With the acceleration of urbanization and the continuous growth of population, the problem of water shortage and uneven distribution has become more and more prominent. In arid and semi-arid areas, this challenge is particularly severe, and the traditional single water source scheduling model can no longer meet the growing demand. As an effective strategy to solve this problem, multi-water source joint scheduling has received widespread attention in recent years.

[0003] Multi-source joint dispatching involves the integration of various types of water sources, such as surface water, groundwater, recycled water, and desalinated water. Each water source has its own unique water quality, water quantity characteristics, cost structure, and supply stability. In the context of flood control water level management, how to efficiently coordinate these resources to achieve optimal economic and environmental benefits has become a key area of ​​current research.

[0004] At present, the joint dispatch of multiple water sources mainly relies on traditional water conservancy engineering facilities and rule-based control systems. For example, in some river basins, a series of cascade reservoirs are built to regulate runoff to achieve multiple functions such as power generation, flood control, and irrigation. Especially during the flood season, the reasonable setting and adjustment of flood limit water levels can effectively reduce flood risks while ensuring the water storage capacity of reservoirs; or in urban water supply systems, according to different water sources and water quality conditions, a simple rotation method or a proportional mixed supply is adopted. However, these traditional methods have certain limitations:

[0005] 1. Lack of flexibility: Traditional scheduling methods are often based on fixed rules or models, which are difficult to adapt to the complex and changeable natural conditions during the flood season and sudden flood events.

[0006] 2. The phenomenon of information islands is serious: data exchange between water sources is not smooth, resulting in the inability to fully and real-time consider all available resources when making decisions on water level management during the flood season, affecting the accuracy and efficiency of scheduling.

[0007] 3. Limited degree of optimization: In flood season water level management, traditional methods fail to fully consider multi-objective optimization, such as how to achieve a balance between maximizing economic benefits and minimizing environmental impact while ensuring flood control safety.

[0008] 4. Low level of automation: Most operations still require manual intervention, especially in emergency situations during flood season. The real-time response capability is insufficient, which affects the timeliness and effectiveness of dispatching decisions. Summary of the invention

[0009] In view of the problems in the related art, the present invention proposes a multi-water source joint scheduling method and system to overcome the above-mentioned technical problems existing in the existing related art.

[0010] To this end, the specific technical solution adopted by the present invention is as follows:

[0011] According to one aspect of the present invention, a multi-water source joint scheduling method is provided, the method comprising the following steps:

[0012] S1. Use machine learning algorithms to analyze the historical water level data and real-time water level data of each reservoir, set the flood limit water level target, and calculate the water level deviation value between the real-time water level data of each reservoir and the flood limit water level target;

[0013] S2. Combine the water level deviation value of each reservoir with the weather data in the preset time period, use the reinforcement learning algorithm to predict the water release volume of each reservoir, and adjust the predicted water release volume based on the real-time water level change trend;

[0014] S3. According to the characteristics of different flood season stages, a coordinated control mechanism between reservoirs is established, the gate flow parameters of each reservoir are calculated through a time series algorithm, and a reservoir gate control model is constructed based on the gate flow parameters;

[0015] S4. Use deep learning technology to predict the water level changes of each reservoir, optimize the flood limit water level target of each reservoir based on the prediction results, and generate gate control instructions in combination with the reservoir gate control model;

[0016] S5. Input the generated gate control instruction into the gate controller, and execute water source scheduling between reservoirs in combination with pre-set scheduling rules.

[0017] Optionally, using a machine learning algorithm to analyze historical water level data and real-time water level data of each reservoir, setting a flood limit water level target, and calculating a water level deviation value between the real-time water level data of each reservoir and the flood limit water level target includes the following steps:

[0018] S11, collecting historical water level data and real-time water level data from each reservoir, and preprocessing the historical water level data and the real-time water level data;

[0019] S12. Construct a random forest regression model based on a machine learning algorithm, and use the preprocessed historical water level data to train the random forest regression model to identify the trend of water level changes;

[0020] S13, inputting the preprocessed real-time water level data into the trained random forest regression model, and outputting the water level prediction result;

[0021] S14. Based on the output water level prediction results and in combination with the water level distribution characteristics of historical flood periods and non-flood periods, set the flood limit water level target;

[0022] S15, comparing the preprocessed real-time water level data with the set flood limit water level target, calculating the difference between the real-time water level data of each reservoir and the flood limit water level target at each time point, and obtaining the water level deviation value.

[0023] Optionally, according to the characteristics of different flood season stages, a coordinated control mechanism between reservoirs is constructed, the gate flow parameters of each reservoir are calculated by a time series algorithm, and a reservoir gate control model is constructed based on the gate flow parameters, including the following steps:

[0024] S31, dividing the flood season into different stages according to the historical water level data and the real-time water level data, combined with the weather data in a preset time period, and identifying the characteristic data of each stage;

[0025] S32. Based on the characteristic data of different flood season stages, analyze the correlation between reservoirs and establish a coordinated control mechanism between reservoirs;

[0026] S33, using a time series algorithm, combined with weather data within a preset time period and the established reservoir collaborative control mechanism, predicting the water level change trend of each reservoir within a preset time period, and calculating the gate flow parameters of each reservoir based on the water level change trend and reservoir operation rules;

[0027] S34. Analyze the parameters of the water flow according to the calculated gate flow parameters to obtain the gate opening data, and construct a reservoir gate control model based on the mathematical relationship between the gate flow parameters and the opening data.

[0028] Optionally, based on the characteristic data of different flood season stages, the correlation between reservoirs is analyzed and a coordinated control mechanism between reservoirs is established, which includes the following steps:

[0029] S321. Based on the characteristic data of different flood season stages, a reservoir network topology diagram including each reservoir node and corresponding connecting pipelines is constructed;

[0030] S322, evaluating the importance of the reservoir nodes and the connecting pipelines in the reservoir network topology diagram, and selecting the remaining reservoir nodes in order of importance of the reservoir nodes;

[0031] S323, combining the retained reservoir nodes with the preset screening rules, screening out the reservoir nodes that serve as water sources and the reservoir nodes that serve as water receivers;

[0032] S324, according to the reservoir network topology diagram, taking the importance of each connecting pipeline as a weight, to form an undirected weighted graph of the reservoir network;

[0033] S325, based on the undirected weighted graph of the reservoir network, calculating the shortest path data set between the retained reservoir nodes and the receiving reservoir nodes and between the retained receiving reservoir node pairs;

[0034] S326, based on the calculated shortest path data set, construct a connection graph including reservoir nodes and corresponding shortest paths, and obtain a minimum spanning tree for the connection graph to obtain an initial collaborative control mechanism;

[0035] S327. Use the water flow simulation model to verify the initial collaborative control mechanism, and obtain the collaborative control mechanism between reservoirs based on the verification results.

[0036] Optionally, based on the undirected weighted graph of the reservoir network, calculating the shortest path data set between the retained reservoir nodes and the receiving reservoir nodes and between the retained receiving reservoir node pairs comprises the following steps:

[0037] S3251. Based on the undirected weighted graph of the reservoir network, determine the starting reservoir node set and the ending reservoir node set, the starting reservoir node set includes the remaining reservoir nodes and the corresponding receiving reservoir nodes, and the ending reservoir node set includes the remaining receiving reservoir nodes and the corresponding receiving reservoir nodes;

[0038] S3252, initializing the total weight value of the shortest path from the starting node to the current node, and clearing the total weight value of the shortest path to zero;

[0039] S3253, initialize the predecessor node pointer of each reservoir node, set the predecessor pointer of the starting reservoir node to null, set the predecessor pointers of the remaining reservoir nodes to preset values, and clear the shortest path calculation result set to store the calculated shortest path;

[0040] S3254, traversing the undirected weighted graph of the reservoir network, searching for all paths from the starting reservoir node to the ending reservoir node according to the reservoir nodes connected by the path, and accumulating the path weights from the starting node to the current node;

[0041] S3255. For each path to the terminal node, record the corresponding cumulative weight, select the path with the smallest cumulative weight as the shortest path, and update the predecessor node pointer of each node on the shortest path;

[0042] S3256, using an iterative algorithm, by traversing and updating the shortest path weights and predecessor pointer states of the reservoir nodes, until the shortest path from each reservoir node in the starting reservoir node set to each reservoir node in the ending reservoir node set is found;

[0043] S3257. Starting from each node in the termination node set, the path is backtracked according to the predecessor node pointer, and the shortest path is gradually constructed until the predecessor pointer is empty, and the shortest path obtained by backtracking is saved in the shortest path data set.

[0044] Optionally, the reservoir gate control model is formulated as:

[0045]

[0046] In the formula, h represents the gate opening value; t represents the time; K p represents proportional gain; e represents error signal; K i represents the integral gain; γ represents every moment in the continuous time period from time 0 to time t; K d Represents the differential gain.

[0047] Optionally, using deep learning technology to predict the water level changes of each reservoir, optimizing the flood limit water level target of each reservoir based on the prediction results, and generating gate control instructions in combination with the reservoir gate control model include the following steps:

[0048] S41. Based on each reservoir, use deep learning technology to build a long short-term memory network model;

[0049] S42, obtaining and preprocessing the historical opening data of the reservoir gate, using the preprocessed historical water level data as input features and the preprocessed historical opening data as output targets, and training the long short-term memory network model;

[0050] S43, inputting the preprocessed real-time water level data into the trained long short-term memory network model to obtain the water level prediction value, and identifying the water level risk point in combination with the set water level threshold;

[0051] S44, based on the identified water level risk points, optimizing the flood limit water level targets of each reservoir using an optimization algorithm;

[0052] S45, inputting the optimized flood limit water level target into the reservoir gate control model, and combining it with the water level prediction value to obtain the gate opening data;

[0053] S46. Combining the optimized flood limit water level target with the obtained gate opening data, a gate control instruction is generated using a PID controller algorithm.

[0054] Optionally, based on the identified water level risk points, optimizing the flood limit water level targets of each reservoir using an optimization algorithm includes the following steps:

[0055] S441, using an optimization algorithm, initializing the flood limit water level target as a vertex, and setting the parameters of the optimization algorithm;

[0056] S442, defining an objective function according to the identified water level risk points and historical water level data, and using the objective function to calculate the objective function values ​​of all vertices;

[0057] S443, based on the objective function values ​​of all vertices, select the best vertex, the worst vertex and the suboptimal vertex, and calculate the central vertex using the positions of the best and suboptimal points;

[0058] S444, performing a reflection operation using the position of the central vertex and the position of the worst vertex to calculate a reflected vertex, and comparing the objective function value of the reflected vertex with the objective function value of the optimal vertex;

[0059] S445. Based on the comparison result between the reflection vertex and the optimal vertex, update the vertex, and repeat steps S442 to S444 until a preset number of iterations is reached to obtain the optimal flood limit water level target.

[0060] Optionally, the formula for reflecting vertices is:

[0061] D r =D c +κ(D c -D w )

[0062] Where D r Represents the reflection vertex; D c represents the central vertex; κ represents the reflection factor; D w Indicates the worst vertex.

[0063] According to another aspect of the present invention, a multi-water source joint dispatching system is also provided, which includes a water level analysis and target setting module, a water release prediction and adjustment module, a mechanism construction and flow calculation module, a water level prediction and instruction generation module, and a gate control and water source dispatching module;

[0064] The water level analysis and target setting module is used to use machine learning algorithms to analyze the historical water level data and real-time water level data of each reservoir, set flood limit water level targets, and calculate the water level deviation value between the real-time water level data of each reservoir and the flood limit water level target;

[0065] The water release prediction and adjustment module is used to combine the water level deviation value of each reservoir with the weather data in the preset time period, use the reinforcement learning algorithm to predict the water release volume of each reservoir, and adjust the predicted water release volume based on the real-time water level change trend;

[0066] The mechanism construction and flow calculation module is used to build a coordinated control mechanism among reservoirs according to the characteristics of different flood season stages, calculate the gate flow parameters of each reservoir through a time series algorithm, and build a reservoir gate control model based on the gate flow parameters;

[0067] The water level prediction and command generation module is used to predict the water level changes of each reservoir using deep learning technology, optimize the flood limit water level target of each reservoir based on the prediction results, and generate gate control commands in combination with the reservoir gate control model;

[0068] The gate control and water source scheduling module is used to input the generated gate control instructions into the gate controller and execute water source scheduling between reservoirs in combination with pre-set scheduling rules.

[0069] The beneficial effects of the present invention are:

[0070] 1. The present invention achieves maximum water resource storage while meeting flood control needs by calculating the water level deviation value; at the same time, the water release volume is predicted by using a reinforcement learning algorithm, and is dynamically adjusted according to the real-time water level change trend, thereby avoiding unnecessary waste of water resources and reducing the risk of isolated operation. This not only improves the resource utilization efficiency of a single reservoir, but also promotes the coordinated operation of multiple reservoirs.

[0071] 2. The present invention predicts the water level changes of each reservoir through deep learning technology, which greatly improves the ability to foresee future hydrological conditions and makes scheduling decisions more scientific and reasonable. At the same time, the prediction results are dynamically adjusted according to the real-time water level change trend, which improves the prediction accuracy and response speed, provides a solid foundation for coping with rapidly changing hydrological conditions, and enhances the flexibility and adaptability of the system.

[0072] 3. The present invention realizes information sharing and linkage scheduling by constructing a collaborative control mechanism between reservoirs, ensuring that the scheduling of each reservoir at different stages is more coordinated and consistent, and further optimizing resource allocation; the generated intelligent gate control instructions are input into the gate controller, and the scheduling is executed in combination with the pre-set scheduling rules, which not only improves the resilience and reliability of the system, but also promotes sustainable development in the region. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 is a flow chart of a multi-water source joint scheduling method according to an embodiment of the present invention;

[0074] Figure 2 It is a principle block diagram of a multi-water source joint scheduling system according to an embodiment of the present invention.

[0075] In the figure:

[0076] 1. Water level analysis and target setting module; 2. Water release prediction and adjustment module; 3. Mechanism construction and flow calculation module; 4. Water level prediction and instruction generation module; 5. Gate control and water source scheduling module. DETAILED DESCRIPTION

[0077] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention and are mainly used to illustrate the embodiments. They can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these contents, ordinary technicians in this field should be able to understand other possible implementation methods and advantages of the present invention.

[0078] According to an embodiment of the present invention, a multi-water source joint scheduling method and system are provided.

[0079] The present invention is further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to the multi-water source joint scheduling method of an embodiment of the present invention, the method includes the following steps:

[0080] S1. Use machine learning algorithms to analyze the historical water level data and real-time water level data of each reservoir, set flood limit water level targets, and calculate the water level deviation value between the real-time water level data of each reservoir and the flood limit water level target.

[0081] Preferably, using a machine learning algorithm to analyze the historical water level data and real-time water level data of each reservoir, setting a flood limit water level target, and calculating the water level deviation value between the real-time water level data of each reservoir and the flood limit water level target includes the following steps:

[0082] S11. Collect historical water level data and real-time water level data from each reservoir, and pre-process the historical water level data and real-time water level data.

[0083] It should be noted that the steps for preprocessing historical water level data and real-time water level data are as follows:

[0084] Step 1: Data cleaning

[0085] Outlier detection: Use statistical methods (such as the 3σ principle) or machine learning algorithms (such as isolation forests) to detect and remove outliers in historical and real-time water level data.

[0086] Missing value processing: For missing values ​​in historical water level data and real-time water level data, interpolation methods (such as linear interpolation, polynomial interpolation) or estimation based on historical trends are used.

[0087] Deletion of duplicate values: Directly delete duplicate records in historical water level data and real-time water level data.

[0088] Step 2: Data formatting

[0089] Timestamp unification: Ensure that the timestamp format of all data is consistent, such as converting to UTC time or local time.

[0090] Data Unit Standardization: Convert water level data to the same unit, such as meters or feet.

[0091] Data format conversion: Convert raw data into a data format suitable for subsequent analysis, such as CSV, Excel, or database table.

[0092] S12. A random forest regression model is constructed based on the machine learning algorithm, and the preprocessed historical water level data is used to train the random forest regression model to identify the trend of water level changes.

[0093] It should be noted that the trained random forest regression model is used to predict the water level in the future, and the prediction results are visualized and compared with the historical water level data to intuitively show the trend of water level changes.

[0094] S13, input the preprocessed real-time water level data into the trained random forest regression model, and output the water level prediction result.

[0095] S14. Based on the output water level prediction results and combined with the water level distribution characteristics during historical flood periods and non-flood periods, set the flood limit water level target.

[0096] It should be noted that based on the output water level forecast results and combined with the water level distribution characteristics of historical flood periods and non-flood periods, the steps for setting the flood limit water level target are as follows:

[0097] Step 1: Analyze water level distribution characteristics

[0098] Statistics on the water level range, average value, maximum value, etc. during historical flood periods are collected to understand the water level characteristics during flood periods. Similarly, statistics on the water level range, average value, etc. during non-flood periods are collected to compare the characteristics of flood periods. Analysis of water level change trends over time, including seasonal changes, long-term trends, etc.

[0099] Step 2: Set flood limit water level target

[0100] Based on the output water level forecast results, a preliminary flood limit water level target is set, and the preliminary flood limit water level target is adjusted according to the water level distribution characteristics of historical flood periods and non-flood periods. If the water level in historical flood periods often exceeds a certain threshold, then this threshold can be used as a reference point for the flood limit water level. For example,

[0101] The water level during historical flood periods ranges from 12.0 metres to 13.5 metres.

[0102] The water level during non-flood season ranges between 10.0 metres and 11.5 metres.

[0103] Water levels vary seasonally, with higher levels in summer and autumn and lower levels in winter and spring.

[0104] Based on this information, a preliminary flood limit water level target can be set, such as 12.0 meters, and then adjusted based on historical flood events to ultimately determine a reasonable flood limit water level target.

[0105] S15, comparing the preprocessed real-time water level data with the set flood limit water level target, calculating the difference between the real-time water level data of each reservoir and the flood limit water level target at each time point, and obtaining the water level deviation value.

[0106] It should be noted that the calculation formula for the water level deviation value is:

[0107] Z=F m -F n

[0108] Where, Z represents the water level deviation value; F m Indicates real-time water level data; F n Indicates the flood limit water level target.

[0109] S2. Combine the water level deviation value of each reservoir with the weather data in the preset time period, use the reinforcement learning algorithm to predict the water release volume of each reservoir, and adjust the predicted water release volume based on the real-time water level change trend.

[0110] It should be noted that the steps of predicting the water release of each reservoir by combining the water level deviation value of each reservoir with the weather data in the preset time period and using the reinforcement learning algorithm to adjust the predicted water release based on the real-time water level change trend are as follows:

[0111] Step 1: Reinforcement Learning Algorithm Design

[0112] Define the state space, including the current water level deviation value, weather data, etc.; define the action space, that is, the possible range of water release, which can be discrete (such as 0, 100, 200 cubic meters per hour) or continuous; design a reward function to reflect the relationship between water release and reservoir safety, downstream flood control, water resource utilization, etc. For example, when the water level deviation value approaches or exceeds the flood limit water level, a negative reward is given; when the water release meets the downstream water demand and the reservoir is safe, a positive reward is given; use a neural network as a policy network, input the data of the state space, and output the water release in the action space.

[0113] Step 2: Train the reinforcement learning model

[0114] The historical water release data, water level deviation values ​​and weather data are preprocessed to form a training data set. The training data set is used to train the reinforcement learning model and optimize the policy network so that it can select the optimal water release amount according to the current status.

[0115] Step 3: Real-time prediction and adjustment

[0116] The real-time water level deviation value and weather data are input into the trained reinforcement learning model, and the model outputs the predicted water release volume; the predicted water release volume is adjusted according to the real-time water level change trend. For example, when the water level rises rapidly, the water release volume may need to be increased to lower the water level; when the water level is stable or falling, the water release volume can be appropriately reduced.

[0117] For example, at 18:00, based on the real-time water level trend and weather data, the model predicted a release rate of 180 cubic meters per hour. However, considering that the water level was rising rapidly, managers decided to adjust the predicted release rate to 170 cubic meters per hour. At 24:00, as the water level had exceeded the flood limit, the model predicted the need to increase the release rate to 200 cubic meters per hour. Based on actual conditions, managers further increased the release rate to 220 cubic meters per hour to ensure reservoir safety.

[0118] S3. According to the characteristics of different flood season stages, a coordinated control mechanism between reservoirs is established, the gate flow parameters of each reservoir are calculated through a time series algorithm, and a reservoir gate control model is constructed based on the gate flow parameters.

[0119] Preferably, according to the characteristics of different stages of the flood season, a coordinated control mechanism between reservoirs is constructed, the gate flow parameters of each reservoir are calculated by a time series algorithm, and a reservoir gate control model is constructed based on the gate flow parameters, including the following steps:

[0120] S31. Based on the historical water level data and the real-time water level data, combined with the weather data in a preset time period, the flood season is divided into different stages, and the characteristic data of each stage is identified.

[0121] It should be noted that, based on historical water level data and real-time water level data, combined with weather data within a preset time period, the flood season is divided into different stages, as shown in Table 1, including:

[0122] Spring flood season: The temperature rises in spring, and seasonal snow or glaciers melt, providing water for rivers and forming spring floods; during this stage, the water level gradually rises, but the flow changes are relatively small and lasts for a short time.

[0123] Summer flood season (summer flood season): There is abundant precipitation in summer, and rivers rely mainly on atmospheric precipitation for replenishment, resulting in summer floods. During this stage, the water level rises rapidly, the flow rate changes greatly, and the duration is relatively long, making it the focus of flood prevention work.

[0124] Autumn flood season: In autumn, some areas have more precipitation, or the meltwater from mountain ice and snow continues to replenish, forming autumn floods; during this stage, the water level changes are between the spring floods and the summer floods, and the flow is relatively small, but it may also cause certain flood disasters.

[0125] Ice flood season: the period of rising water levels in winter and spring due to ice blockage and thawing of rivers. This stage mainly occurs in rivers with a freezing period. The water level changes are affected by ice, which may cause dangerous situations such as ice jams and ice dams.

[0126] Table 1: Schematic diagram of flood season

[0127]

[0128] By comparing historical water level data with real-time water level data and combining it with weather data within a preset time period, we can find that:

[0129] From March to May, as spring temperatures rise and snow melts, water levels gradually rise and rainfall gradually increases, meeting the characteristics of the spring flood season.

[0130] During the period from June to August, rainfall increased significantly, water levels rose rapidly, and temperatures were relatively high, which is consistent with the characteristics of the summer flood season.

[0131] During September to November, rainfall decreased, but the water level remained high and the temperature gradually dropped, which is consistent with the characteristics of the autumn flood season.

[0132] From December to February of the following year, although the rainfall is relatively small, the water level may fluctuate due to the influence of ice and icicles, which is consistent with the characteristics of the ice flood season.

[0133] S32. Based on the characteristic data of different flood season stages, analyze the correlation between reservoirs and establish a coordinated control mechanism among reservoirs.

[0134] Preferably, based on the characteristic data of different flood season stages, the correlation between reservoirs is analyzed and a coordinated control mechanism between reservoirs is established, which includes the following steps:

[0135] S321. Based on the characteristic data of different flood season stages, a reservoir network topology diagram including each reservoir node and the corresponding connecting pipelines is constructed.

[0136] It should be noted that the steps for constructing a reservoir network topology diagram containing each reservoir node and the corresponding connecting pipelines based on the characteristic data of different flood season stages are as follows:

[0137] Step 1: Determine the reservoir node

[0138] Identify all reservoir nodes in the system. These can be large, medium, or small reservoirs, depending on the size and complexity of the system.

[0139] Step 2: Determine the connection pipeline

[0140] Determine the connecting pipes between each reservoir node, which can be water pipes, spillways or other forms of connections.

[0141] Step 3: Build a topology diagram

[0142] Use graphic software (such as Visio, AutoCAD, etc.) or online drawing tools (such as Lucidchart, Draw.io, etc.) to construct a reservoir network topology diagram. In the diagram, each reservoir node can be represented by a circle, rectangle or other shape, and the connecting pipes can be represented by lines or arrows; according to the characteristic data of different stages of the flood season, the water level, flow and other information of each node are marked in the diagram to reflect the changes in different stages of the flood season.

[0143] In a specific embodiment, it is assumed that there are three reservoir nodes: reservoir A, reservoir B and reservoir C. Reservoir A is connected to reservoir B through pipeline 1, reservoir B is connected to reservoir C through pipeline 2, and reservoir A is also directly connected to reservoir C through pipeline 3 (as a backup connection).

[0144] At the beginning of the flood season (such as the spring flood season), the water level of each reservoir is low and the flow is small; when constructing or updating the reservoir network topology diagram, the water level and flow values ​​of these nodes can be marked as lower levels accordingly.

[0145] As the flood season progresses (such as entering the summer flood season), rainfall increases, the water levels of various reservoirs gradually rise, and the flow rate also increases accordingly; in order to accurately reflect this change, it is necessary to timely update the water level and flow information in the reservoir network topology diagram.

[0146] At the end of the flood season (such as the autumn flood season or the ice flood season), the rainfall decreases, but the water levels of some reservoirs may still be high due to the influence of previous rainfall and snowmelt; the water level and flow information in the reservoir network topology diagram can continue to be updated to ensure that it can accurately reflect the actual situation at the end of the flood season.

[0147] S322. Evaluate the importance of the reservoir nodes and connecting pipelines in the reservoir network topology diagram, and select the remaining reservoir nodes in order of importance.

[0148] It should be noted that importance assessment includes: node degree assessment, betweenness assessment, aggregation degree assessment and actual data support.

[0149] Node degree evaluation: Node degree refers to the number of pipelines directly connected to the node. The higher the node degree, the stronger its connectivity in the network, and the greater the impact of its failure on the network.

[0150] Betweenness evaluation: Betweenness refers to the number of paths that pass through the node among all the shortest paths in the network. The higher the betweenness of a node, the more obvious its "bridge" role in the network. Its failure may lead to network segmentation or loss of function.

[0151] Convergence evaluation: Convergence is a comprehensive indicator to measure the importance of a node in the network, which takes into account multiple factors such as the degree and location of the node. By shrinking the node and observing the changes in the network's convergence, the importance of the node can be evaluated.

[0152] Actual data support: Combining historical data, real-time monitoring data and prediction models, quantitative analysis of key indicators such as water level, flow, stability, etc. of nodes is carried out to further support importance assessment.

[0153] In a specific embodiment, assuming that there is a reservoir network including reservoir A, reservoir B, reservoir C and reservoir D, and pipelines connecting them, the process of evaluating and ranking the importance of nodes is as follows:

[0154] Step 1: Initial Assessment

[0155] Reservoir A: It has a high degree value, is connected to multiple reservoirs, has a high betweenness, and is a key node in the network.

[0156] Reservoir B: The degree value is medium, but the betweenness is high, indicating that it plays a key role in some shortest paths.

[0157] Reservoir C: has a lower value, but connects to an important water source or drainage area and therefore has specific functional importance.

[0158] Reservoir D: The degree value is low, and its position in the network is relatively marginal, and the betweenness is also low.

[0159] Step 2: Polymerization Degree Evaluation and Verification

[0160] By shrinking the nodes and observing the changes in network aggregation, it was found that the network aggregation decreased most significantly after shrinking reservoir A, verifying its importance.

[0161] After shrinking reservoir B, the network aggregation degree also decreases to a certain extent, but the magnitude is smaller than that of reservoir A.

[0162] The shrinking reservoir C has a smaller impact on the network aggregation, but it still needs to be retained considering its specific functional importance.

[0163] The shrinking reservoir D has the least impact on the network aggregation.

[0164] Step 3: Comprehensive sorting and selection

[0165] Taking into account node degree, betweenness, aggregation degree and actual data support, the reservoir nodes are ranked from high to low in importance as follows: Reservoir A>Reservoir B>Reservoir C>Reservoir D.

[0166] Reservoirs A and B should be retained as a priority to ensure the stability and key functions of the network; depending on actual conditions and needs, reservoir C can be considered to maintain specific functions, while reservoir D may be used as an alternative or optimization object.

[0167] S323. Combine the retained reservoir nodes with the preset screening rules to screen out the reservoir nodes that serve as water sources and the reservoir nodes that serve as water receivers.

[0168] It should be noted that the screening rules include water source node screening rules and water receiving node screening rules.

[0169] Water source node screening rules: The reservoir has a large capacity and can store enough rainwater or upstream water; the water quality is excellent and meets the standards for drinking water or irrigation water; the geographical location is high, which is convenient for supplying water to downstream reservoirs through gravity flow or pumping.

[0170] Water receiving node screening rules: The reservoir capacity is relatively small and is mainly used to store and regulate downstream water; the geographical location is low, which is convenient for receiving water supply from upstream reservoirs; it plays a key regulation and distribution role in the network and can ensure the water demand of downstream areas.

[0171] In a specific embodiment, there are reservoirs A, B and C, and the analysis is as follows:

[0172] (1) Analysis of Reservoir A:

[0173] Capacity: Large, able to store large amounts of rainwater.

[0174] Water quality: Excellent, meeting drinking water standards.

[0175] Geographical location: High, located upstream.

[0176] Conclusion: According to the screening rules, Reservoir A meets all the conditions of the water source node and is therefore determined to be the water source node.

[0177] (2) Analysis of Reservoir B:

[0178] Capacity: Medium, able to store a certain amount of rainwater.

[0179] Water quality: Good, but slightly inferior to Reservoir A.

[0180] Geographical location: Located between Reservoir A and Reservoir C, it plays a role of transit and regulation.

[0181] Conclusion: Although the water quality and capacity of Reservoir B are good, its geographical location makes it more suitable as a water receiving node, receiving water from Reservoir A and supplying water to the downstream Reservoir C.

[0182] (3) Reservoir C analysis:

[0183] Capacity: Small, mainly used to regulate downstream water.

[0184] Water quality: Affected by upstream reservoirs, but generally good.

[0185] Geographical location: Low, located downstream.

[0186] Conclusion: According to the screening rules, reservoir C meets all the conditions of the water receiving node and is therefore determined to be the water receiving node.

[0187] Based on the above analysis and screening rules, the following screening results were obtained:

[0188] Water source node: Reservoir A.

[0189] Water receiving nodes: Reservoir B (transit and regulation), Reservoir C (terminal water receiving).

[0190] S324. According to the reservoir network topology diagram, the importance of each connecting pipeline is used as a weight to form an undirected weighted graph of the reservoir network.

[0191] S325. Based on the undirected weighted graph of the reservoir network, calculate the shortest path data set between the retained reservoir nodes and the receiving reservoir nodes and between the retained receiving reservoir node pairs.

[0192] Preferably, based on the undirected weighted graph of the reservoir network, calculating the shortest path data set between the retained reservoir nodes and the receiving reservoir nodes and between the retained receiving reservoir node pairs comprises the following steps:

[0193] S3251. Based on the undirected weighted graph of the reservoir network, determine the starting reservoir node set and the ending reservoir node set, the starting reservoir node set includes the retained reservoir nodes and the corresponding receiving reservoir nodes, and the ending reservoir node set includes the retained receiving reservoir nodes and the corresponding receiving reservoir nodes.

[0194] S3252: Initialize the total weight value of the shortest path from the starting node to the current node, and clear the total weight value of the shortest path to zero.

[0195] S3253, initialize the predecessor node pointer of each reservoir node, set the predecessor pointer of the starting reservoir node to null, set the predecessor pointers of the remaining reservoir nodes to preset values, and clear the shortest path calculation result set to store the calculated shortest path.

[0196] S3254, traverse the undirected weighted graph of the reservoir network, search for all paths from the starting reservoir node to the ending reservoir node according to the reservoir nodes connected by the path, and accumulate the path weights from the starting node to the current node.

[0197] S3255. For each path to the terminal node, the corresponding cumulative weight is recorded, and the path with the smallest cumulative weight is selected as the shortest path, and the predecessor node pointer of each node on the shortest path is updated.

[0198] S3256. Using an iterative algorithm, by traversing and updating the shortest path weights and predecessor pointer states of the reservoir nodes, the shortest path from each reservoir node in the starting reservoir node set to each reservoir node in the ending reservoir node set is found.

[0199] S3257. Starting from each node in the termination node set, the path is backtracked according to the predecessor node pointer, and the shortest path is gradually constructed until the predecessor pointer is empty, and the shortest path obtained by backtracking is saved in the shortest path data set.

[0200] S326. Based on the calculated shortest path data set, a connection graph including reservoir nodes and corresponding shortest paths is constructed, and a minimum spanning tree is obtained for the connection graph to obtain an initial collaborative control mechanism.

[0201] S327. Use the water flow simulation model to verify the initial collaborative control mechanism, and obtain the collaborative control mechanism between reservoirs based on the verification results.

[0202] S33. Utilize the time series algorithm, combine the weather data within the preset time period with the established collaborative control mechanism between reservoirs, predict the water level change trend of each reservoir within the preset time period, and calculate the gate flow parameters of each reservoir based on the water level change trend and reservoir operation rules.

[0203] It should be noted that the ARIMA model (autoregressive integrated moving average model) was selected as the time series prediction model using the time series algorithm, and the water level change trend of reservoir A was predicted using the ARIMA model, as follows:

[0204] The water level data of Reservoir A in the past year was collected, cleaned and standardized, and the water level data of Reservoir A was trained using the ARIMA model to obtain model parameters; the weather forecast data for the next week was input, and the water level change trend of Reservoir A was predicted using the trained ARIMA model; the prediction results show that the water level of Reservoir A will gradually rise in the next week and reach a peak on the fourth day. According to the operating rules and water level change trend of Reservoir A, it is calculated that the gate needs to be opened on the fourth day to release water to lower the water level; the specific gate flow parameters include the gate opening degree of 50%, the opening time is from 8 am to 4 pm on the fourth day, and the flow size is 10 cubic meters per second.

[0205] In addition, reservoir operation rules refer to a series of rules, regulations and operational requirements formulated in the operation and management of reservoirs to ensure the safe and efficient operation of reservoirs and give full play to their comprehensive benefits, including: daily reservoir operation rules, reservoir safety management rules, water resources allocation rules and other operation rules.

[0206] S34. Analyze the parameters of the water flow according to the calculated gate flow parameters to obtain the gate opening data, and construct a reservoir gate control model based on the mathematical relationship between the gate flow parameters and the opening data.

[0207] Preferably, the formula of the reservoir gate control model is:

[0208]

[0209] In the formula, h represents the gate opening value; t represents the time; K p represents proportional gain; e represents error signal; K i represents the integral gain; γ represents every moment in the continuous time period from time 0 to time t; K d Represents the differential gain.

[0210] S4. Use deep learning technology to predict the water level changes of each reservoir, optimize the flood limit water level target of each reservoir based on the prediction results, and generate gate control instructions in combination with the reservoir gate control model.

[0211] Preferably, using deep learning technology to predict the water level changes of each reservoir, optimizing the flood limit water level target of each reservoir based on the prediction results, and generating gate control instructions in combination with the reservoir gate control model include the following steps:

[0212] S41. Based on each reservoir, deep learning technology is used to construct a long-short-term memory network model.

[0213] S42. Acquire and preprocess the historical opening data of the reservoir gate, use the preprocessed historical water level data as input features, and the preprocessed historical opening data as output targets to train the long short-term memory network model.

[0214] S43, input the preprocessed real-time water level data into the trained long short-term memory network model to obtain the water level prediction value, and identify the water level risk point in combination with the set water level threshold.

[0215] S44. Based on the identified water level risk points, the flood limit water level targets of each reservoir are optimized using an optimization algorithm.

[0216] Preferably, based on the identified water level risk points, optimizing the flood limit water level targets of each reservoir using an optimization algorithm includes the following steps:

[0217] S441. Using an optimization algorithm, initialize the flood limit water level target as a vertex, and set the parameters of the optimization algorithm.

[0218] S442. Define an objective function based on the identified water level risk points and historical water level data, and use the objective function to calculate the objective function values ​​of all vertices.

[0219] S443. Based on the objective function values ​​of all vertices, the optimal vertex, the worst vertex and the suboptimal vertex are selected, and the central vertex is calculated using the positions of the optimal and suboptimal vertex.

[0220] S444. Perform a reflection operation using the position of the central vertex and the position of the worst vertex to calculate a reflected vertex, and compare the objective function value of the reflected vertex with the objective function value of the optimal vertex.

[0221] Preferably, the formula for reflecting the vertex is:

[0222] D r =D c +κ(D c -D w );

[0223] Where D r Represents the reflection vertex; D c represents the central vertex; κ represents the reflection factor; D w Indicates the worst vertex.

[0224] It should be noted that the reflection factor is generally taken as 1.

[0225] S445. Based on the comparison result between the reflection vertex and the optimal vertex, update the vertex, and repeat steps S442 to S444 until a preset number of iterations is reached to obtain the optimal flood limit water level target.

[0226] It should be noted that the comparison results between the reflection vertex and the optimal vertex include:

[0227] (1) If the objective function value of the reflection vertex is greater than the objective function value of the optimal vertex, it means that the reflection direction effect is poor and compression operation is required to calculate the compressed vertex. The formula is:

[0228] D k =D c +α(D c -D w )

[0229] Where D k Indicates compressed vertex; D c represents the central vertex; α represents the compression factor, which is generally taken as 0.5; D w Indicates the worst vertex.

[0230] If the objective function value of the compressed vertex is less than the objective function value of the worst vertex, the worst vertex is replaced by the compressed vertex; otherwise, the simplex is reduced (the simplex is formed based on the initialized flood limit water level target).

[0231] (2) If the objective function value of the reflected vertex is less than the objective function value of the optimal vertex, it means that the reflection direction is correct, then further try the expansion operation and calculate the extended vertex. The formula is:

[0232] D q =D c +β(D r -D c )

[0233] Where D q Denotes an extended vertex; D c represents the central vertex; β represents the expansion factor, which is generally taken as 2; D r Represents a reflective vertex.

[0234] If the objective function value of the extended vertex is less than that of the optimal vertex, the worst vertex is replaced by the extended vertex; if the objective function value of the extended vertex is greater than that of the optimal vertex, the worst point is replaced by the reflection point.

[0235] (3) If the objective function value of the optimal vertex is less than the objective function value of the reflection vertex and less than the objective function value of the worst vertex, a contraction operation is performed to calculate the inner contraction vertex. The formula is:

[0236] D g =D c -α(D w -D c )

[0237] Where D g Indicates the inward contraction vertex; D c represents the central vertex; α represents the compression factor; D w Indicates the worst vertex.

[0238] If the objective function value of the inner contraction vertex is less than the objective function value of the worst vertex, replace the worst point with the contraction point; otherwise, replace the worst point with the reflection point

[0239] S45. Input the optimized flood limit water level target into the reservoir gate control model, and combine it with the water level prediction value to obtain the gate opening data.

[0240] S46. Combining the optimized flood limit water level target with the obtained gate opening data, a gate control instruction is generated using a PID controller algorithm.

[0241] S5. Input the generated gate control instruction into the gate controller, and execute water source scheduling between reservoirs in combination with pre-set scheduling rules.

[0242] It should be noted that the scheduling rules include:

[0243] The principle of safety ensures that no safety accidents such as flooding and reservoir collapse occur during the scheduling process; the principle of economy maximizes the economic benefits of scheduling and achieves optimal allocation of water resources; the principle of sustainability considers the long-term development of the ecological environment and maintains ecological balance.

[0244] In a specific embodiment, based on the above principles, the following scheduling rules are set:

[0245] When the water level in reservoir A exceeds a certain set threshold, water begins to be released to reservoir B.

[0246] During the water release process, it is necessary to monitor the water level changes of Reservoir A and Reservoir B, as well as the flow conditions downstream in real time.

[0247] Adjust the water release flow according to downstream irrigation, power generation and other needs.

[0248] When the water level of reservoir B reaches the set target, stop releasing water.

[0249] like Figure 2 As shown, according to another embodiment of the present invention, a multi-water source joint scheduling system is also provided, which includes a water level analysis and target setting module 1, a water release prediction and adjustment module 2, a mechanism construction and flow calculation module 3, a water level prediction and instruction generation module 4 and a gate control and water source scheduling module 5;

[0250] Water level analysis and target setting module 1, used to analyze the historical water level data and real-time water level data of each reservoir using machine learning algorithms, set flood limit water level targets, and calculate the water level deviation value between the real-time water level data of each reservoir and the flood limit water level target;

[0251] The water release prediction and adjustment module 2 is used to combine the water level deviation value of each reservoir with the weather data in a preset time period, use the reinforcement learning algorithm to predict the water release volume of each reservoir, and adjust the predicted water release volume based on the real-time water level change trend;

[0252] Mechanism construction and flow calculation module 3 is used to build a coordinated control mechanism among reservoirs according to the characteristics of different flood season stages, calculate the gate flow parameters of each reservoir through a time series algorithm, and build a reservoir gate control model based on the gate flow parameters;

[0253] The water level prediction and instruction generation module 4 is used to predict the water level changes of each reservoir using deep learning technology, optimize the flood limit water level target of each reservoir based on the prediction results, and generate gate control instructions in combination with the reservoir gate control model;

[0254] The gate control and water source scheduling module 5 is used to input the generated gate control instructions into the gate controller, and execute water source scheduling between reservoirs in combination with pre-set scheduling rules.

[0255] In summary, with the help of the above technical scheme of the present invention, by calculating the water level deviation value, it is possible to maximize water resource storage while meeting flood control needs; at the same time, the water release volume is predicted by the reinforcement learning algorithm, and dynamically adjusted according to the real-time water level change trend, avoiding unnecessary waste of water resources and reducing the risk of isolated operation, which not only improves the resource utilization efficiency of a single reservoir, but also promotes the coordinated operation between multiple reservoirs. The water level changes of each reservoir are predicted by deep learning technology, which greatly improves the ability to predict future hydrological conditions and makes the scheduling decision more scientific and reasonable; at the same time, the prediction results are dynamically adjusted according to the real-time water level change trend, which improves the accuracy and response speed of the prediction, provides a solid foundation for responding to rapidly changing hydrological conditions, and enhances the flexibility and adaptability of the system. By constructing a collaborative control mechanism between reservoirs, information sharing and linkage scheduling are realized, ensuring that the scheduling of each reservoir at different stages is more coordinated and consistent, and further optimizing resource allocation; the generated intelligent gate control instructions are input into the gate controller, and the scheduling is performed in combination with the pre-set scheduling rules, which not only improves the resilience and reliability of the system, but also promotes sustainable development in the region.

[0256] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A multi-water source joint scheduling method, characterized in that: The following steps are involved: S1. Use machine learning algorithms to analyze the historical water level data and real-time water level data of each reservoir, set the flood limit water level target, and calculate the water level deviation value between the real-time water level data of each reservoir and the flood limit water level target; S2. Combine the water level deviation value of each reservoir with the weather data in the preset time period, use the reinforcement learning algorithm to predict the water release volume of each reservoir, and adjust the predicted water release volume based on the real-time water level change trend; S3. According to the characteristics of different flood season stages, a coordinated control mechanism between reservoirs is established, the gate flow parameters of each reservoir are calculated through a time series algorithm, and a reservoir gate control model is constructed based on the gate flow parameters; S4. Use deep learning technology to predict the water level changes of each reservoir, optimize the flood limit water level target of each reservoir based on the prediction results, and generate gate control instructions in combination with the reservoir gate control model; S5. Input the generated gate control instruction into the gate controller, and execute water source scheduling between reservoirs in combination with pre-set scheduling rules.

2. The multi-water source joint scheduling method according to claim 1, characterized in that: The S1 comprises the following steps: S11, collecting historical water level data and real-time water level data from each reservoir, and preprocessing the historical water level data and the real-time water level data; S12. Construct a random forest regression model based on a machine learning algorithm, and use the preprocessed historical water level data to train the random forest regression model to identify the trend of water level changes; S13, inputting the preprocessed real-time water level data into the trained random forest regression model, and outputting the water level prediction result; S14. Based on the output water level prediction results and in combination with the water level distribution characteristics of historical flood periods and non-flood periods, set the flood limit water level target; S15, comparing the preprocessed real-time water level data with the set flood limit water level target, calculating the difference between the real-time water level data of each reservoir and the flood limit water level target at each time point, and obtaining the water level deviation value.

3. The multi-water source joint scheduling method according to claim 1, characterized in that: The S3 comprises the following steps: S31, dividing the flood season into different stages according to the historical water level data and the real-time water level data, combined with the weather data in a preset time period, and identifying the characteristic data of each stage; S32. Based on the characteristic data of different flood season stages, analyze the correlation between reservoirs and establish a coordinated control mechanism between reservoirs; S33, using a time series algorithm, combined with weather data within a preset time period and the established reservoir collaborative control mechanism, predicting the water level change trend of each reservoir within a preset time period, and calculating the gate flow parameters of each reservoir based on the water level change trend and reservoir operation rules; S34. Analyze the parameters of the water flow according to the calculated gate flow parameters to obtain the gate opening data, and construct a reservoir gate control model based on the mathematical relationship between the gate flow parameters and the opening data.

4. The multi-water source joint scheduling method according to claim 3 is characterized in that: The S32 comprises the following steps: S321. Based on the characteristic data of different flood season stages, a reservoir network topology diagram including each reservoir node and corresponding connecting pipelines is constructed; S322, evaluating the importance of the reservoir nodes and the connecting pipelines in the reservoir network topology diagram, and selecting the remaining reservoir nodes in order of importance of the reservoir nodes; S323, combining the retained reservoir nodes with the preset screening rules, screening out the reservoir nodes that serve as water sources and the reservoir nodes that serve as water receivers; S324, according to the reservoir network topology diagram, taking the importance of each connecting pipeline as a weight, to form an undirected weighted graph of the reservoir network; S325, based on the undirected weighted graph of the reservoir network, calculating the shortest path data set between the retained reservoir nodes and the receiving reservoir nodes and between the retained receiving reservoir node pairs; S326, based on the calculated shortest path data set, construct a connection graph including reservoir nodes and corresponding shortest paths, and obtain a minimum spanning tree for the connection graph to obtain an initial collaborative control mechanism; S327. Use the water flow simulation model to verify the initial collaborative control mechanism, and obtain the collaborative control mechanism between reservoirs based on the verification results.

5. The multi-water source joint dispatching method according to claim 4 is characterized in that: The S325 comprises the following steps: S3251. Based on the undirected weighted graph of the reservoir network, determine a starting reservoir node set and an ending reservoir node set, wherein the starting reservoir node set includes the remaining reservoir nodes and the corresponding receiving reservoir nodes, and the ending reservoir node set includes the remaining receiving reservoir nodes and the corresponding receiving reservoir nodes; S3252, initializing the total weight value of the shortest path from the starting node to the current node, and clearing the total weight value of the shortest path to zero; S3253, initialize the predecessor node pointer of each reservoir node, set the predecessor pointer of the starting reservoir node to null, set the predecessor pointers of the remaining reservoir nodes to preset values, and clear the shortest path calculation result set to store the calculated shortest path; S3254, traversing the undirected weighted graph of the reservoir network, searching for all paths from the starting reservoir node to the ending reservoir node according to the reservoir nodes connected by the path, and accumulating the path weights from the starting node to the current node; S3255. For each path to the terminal node, record the corresponding cumulative weight, and select the path with the smallest cumulative weight as the shortest path, and update the predecessor node pointer of each node on the shortest path; S3256, using an iterative algorithm, by traversing and updating the shortest path weights and predecessor pointer states of the reservoir nodes, until the shortest path from each reservoir node in the starting reservoir node set to each reservoir node in the ending reservoir node set is found; S3257. Starting from each node in the termination node set, the path is backtracked according to the predecessor node pointer, and the shortest path is gradually constructed until the predecessor pointer is empty, and the shortest path obtained by backtracking is saved in the shortest path data set.

6. The multi-water source joint dispatching method according to claim 3 is characterized in that: The formula of the reservoir gate control model is: In the formula, h represents the gate opening value; t represents the time; K p represents proportional gain; e represents error signal; K i represents the integral gain; γ represents every moment in the continuous time period from time 0 to time t; K d Represents the differential gain.

7. The multi-water source joint scheduling method according to claim 1, characterized in that: The S4 comprises the following steps: S41. Based on each reservoir, use deep learning technology to build a long short-term memory network model; S42, obtaining and preprocessing the historical opening data of the reservoir gate, using the preprocessed historical water level data as input features and the preprocessed historical opening data as output targets, and training the long short-term memory network model; S43, inputting the preprocessed real-time water level data into the trained long short-term memory network model to obtain the water level prediction value, and identifying the water level risk point in combination with the set water level threshold; S44, based on the identified water level risk points, optimizing the flood limit water level targets of each reservoir using an optimization algorithm; S45, inputting the optimized flood limit water level target into the reservoir gate control model, and combining it with the water level prediction value to obtain the gate opening data; S46. Combining the optimized flood limit water level target with the obtained gate opening data, a gate control instruction is generated using a PID controller algorithm.

8. The multi-water source joint dispatching method according to claim 7, characterized in that: The S44 comprises the following steps: S441, using an optimization algorithm, initializing the flood limit water level target as a vertex, and setting the parameters of the optimization algorithm; S442, defining an objective function according to the identified water level risk points and historical water level data, and using the objective function to calculate the objective function values ​​of all vertices; S443, based on the objective function values ​​of all vertices, select the best vertex, the worst vertex and the suboptimal vertex, and calculate the central vertex using the positions of the best and suboptimal points; S444, performing a reflection operation using the position of the central vertex and the position of the worst vertex to calculate a reflected vertex, and comparing the objective function value of the reflected vertex with the objective function value of the optimal vertex; S445. Based on the comparison result between the reflection vertex and the optimal vertex, update the vertex, and repeat steps S442 to S444 until a preset number of iterations is reached to obtain the optimal flood limit water level target.

9. The multi-water source joint dispatching method according to claim 8, characterized in that: The formula for the reflection vertex is: D r =D c +κ(D c -D w ) Where D r Represents the reflection vertex; D c represents the central vertex; κ represents the reflection factor; D w Indicates the worst vertex.

10. A multi-water source joint dispatching system, used to implement the multi-water source joint dispatching method according to any one of claims 1 to 9, characterized in that: The system includes a water level analysis and target setting module, a water release prediction and adjustment module, a mechanism construction and flow calculation module, a water level prediction and instruction generation module, and a gate control and water source scheduling module; The water level analysis and target setting module is used to use machine learning algorithms to analyze the historical water level data and real-time water level data of each reservoir, set flood limit water level targets, and calculate the water level deviation value between the real-time water level data of each reservoir and the flood limit water level target; The water release prediction and adjustment module is used to combine the water level deviation value of each reservoir with the weather data in the preset time period, use the reinforcement learning algorithm to predict the water release volume of each reservoir, and adjust the predicted water release volume based on the real-time water level change trend; The mechanism construction and flow calculation module is used to build a coordinated control mechanism among reservoirs according to the characteristics of different flood season stages, calculate the gate flow parameters of each reservoir through a time series algorithm, and build a reservoir gate control model based on the gate flow parameters; The water level prediction and command generation module is used to predict the water level changes of each reservoir using deep learning technology, optimize the flood limit water level target of each reservoir based on the prediction results, and generate gate control commands in combination with the reservoir gate control model; The gate control and water source scheduling module is used to input the generated gate control instructions into the gate controller and execute water source scheduling between reservoirs in combination with pre-set scheduling rules.

Images