A reservoir scheduling and management system and method

Abstract

The application provides a reservoir scheduling management system and method, relates to the technical field of reservoir scheduling management, and first constructs a reservoir scheduling entity correlation graph containing reservoirs, runoff input, water demand and scheduling rule nodes and having a weighted influence conduction path between the nodes, then dynamically evolves and updates the correlation graph based on hydrological time series data to generate a time-varying scheduling correlation graph; then the time-varying scheduling correlation graph is subjected to multi-agent collaborative reasoning to obtain an influence degree matrix and a scheduling priority sequence, and then a reservoir joint scheduling scheme containing water source allocation, demand response and rule coordination sub-schemes is established according to the influence degree matrix and the scheduling priority sequence; finally, a cascade scheduling instruction set is generated and sent to a reservoir control center to control reservoir gate regulation, unit output adjustment and water delivery flow distribution, so that scientific and efficient scheduling of the reservoir is realized.

Description

Technical Field

[0001] This invention relates to the field of reservoir scheduling and management technology, and more specifically, to a reservoir scheduling and management system and method. Background Technology

[0002] In the field of reservoir scheduling and management, traditional methods have many limitations. Currently, most common reservoir scheduling methods are based on fixed scheduling rules and experience. For example, when formulating scheduling plans, only basic parameters such as the reservoir's capacity and water level are often considered, along with simple historical runoff and water demand data. A comprehensive and in-depth analysis of the complex relationships between the reservoir and its surrounding environment and other related factors is lacking.

[0003] Existing scheduling models typically treat reservoirs, runoff input, and water demand in isolation, failing to construct a holistic model that comprehensively reflects the interactions and transmission relationships among these elements. During scheduling, it is difficult to dynamically adjust based on real-time hydrological changes and fluctuations in water demand, and it cannot accurately and promptly reflect the impact of changes in the state of each element on reservoir scheduling. Furthermore, when facing joint scheduling of multiple reservoirs, the lack of an effective collaborative reasoning mechanism prevents the full consideration of the connections and mutual influences between reservoirs, resulting in scheduling schemes lacking scientific rigor and rationality, hindering the efficient utilization and rational allocation of water resources. Summary of the Invention

[0004] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, an embodiment of the present invention provides a reservoir scheduling and management method, the method comprising:

[0005] Construct a reservoir scheduling entity association graph, which includes reservoir entity nodes, runoff input nodes, water demand nodes, and scheduling rule nodes. Each node establishes an influence transmission path through directed edges, and the influence transmission path is marked with a path influence weight.

[0006] The reservoir scheduling entity association diagram is dynamically updated based on hydrological time series data. The real-time state attributes of each node are updated through the node state propagation algorithm. The path influence weight of the influence transmission path is adjusted according to the real-time state attributes to generate a time-varying scheduling association diagram.

[0007] Multi-agent collaborative reasoning is performed on the time-varying scheduling association graph. The entity association relationship in the time-varying scheduling association graph is deeply reasoned through the reservoir scheduling reasoning model to obtain the influence degree matrix between entities and the scheduling priority sequence.

[0008] A joint reservoir scheduling scheme is established based on the influence matrix and scheduling priority sequence. The joint reservoir scheduling scheme includes a water source allocation sub-scheme, a demand response sub-scheme, and a rule coordination sub-scheme.

[0009] A cascade scheduling instruction set is generated based on the joint reservoir scheduling scheme, and the cascade scheduling instruction set is sent to the reservoir control center to control the regulation of reservoir gates, the adjustment of unit output, and the distribution of water flow.

[0010] In another aspect, embodiments of the present invention also provide a reservoir scheduling and management system, including a processor and a machine-readable storage medium connected to the processor. The machine-readable storage medium is used to store programs, instructions, or code, and the processor is used to execute the programs, instructions, or code in the machine-readable storage medium to implement the above-described method.

[0011] Based on the above, this embodiment of the invention constructs a reservoir scheduling entity association graph, organically integrating reservoir entity nodes, runoff input nodes, water demand nodes, and scheduling rule nodes. It clarifies the influence transmission paths and path influence weights between each node, dynamically evolves and updates the association graph based on hydrological time-series data, reflecting real-time changes in the state of each node and adjusting path influence weights according to the state. This generates a time-varying scheduling association graph, enabling the scheduling model to adapt to constantly changing hydrological and water demand conditions, improving the timeliness and accuracy of scheduling. Multi-agent collaborative reasoning is performed on the time-varying scheduling association graph, and the reservoir scheduling reasoning model is used to deeply mine the relationships between entities, obtaining an influence degree matrix and scheduling priority sequence. The reservoir joint scheduling scheme established based on the influence degree matrix and scheduling priority sequence comprehensively considers multiple aspects such as water source allocation, demand response, and rule coordination, achieving multi-objective optimal allocation of water resources. Finally, a cascade scheduling instruction set is generated based on the joint scheduling scheme and sent to the reservoir control center, enabling precise control of reservoir gate regulation, unit output adjustment, and water flow distribution, effectively improving the intelligence level and operational efficiency of reservoir scheduling, and ensuring the safe and sustainable use of water resources. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of the execution flow of the reservoir scheduling and management method provided in the embodiment of the present invention.

[0013] Figure 2 This is a schematic diagram of exemplary hardware and software components of the reservoir scheduling and management system provided in an embodiment of the present invention. Detailed Implementation

[0014] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a schematic flowchart of a reservoir scheduling and management method provided in one embodiment of the present invention. The reservoir scheduling and management method will be described in detail below.

[0015] Step S110: Construct a reservoir scheduling entity association graph, which includes reservoir entity nodes, runoff input nodes, water demand nodes, and scheduling rule nodes. Each node establishes an influence transmission path through directed edges, and the influence transmission path is marked with a path influence weight.

[0016] In this embodiment, a cascade reservoir group consisting of three reservoirs within a certain region is taken as the scheduling object, and a reservoir scheduling entity association graph is specifically constructed. Within this cascade reservoir group, there are hydraulic connections between the reservoirs, and the outflow from the upstream reservoir affects the inflow into the downstream reservoir. In this scenario, it is necessary to clarify the specific composition of each type of node, establish the connection relationships between nodes, and assign appropriate initial weights to the influence transmission paths.

[0017] Step S111: Determine the node types of the reservoir scheduling entity association graph. The node types include reservoir entity nodes, runoff input nodes, water demand nodes, and scheduling rule nodes. Reservoir entity nodes are used to represent the physical state information of a single reservoir, runoff input nodes are used to represent the natural water inflow information into the reservoir, water demand nodes are used to represent the user demand information that depends on the reservoir for water supply, and scheduling rule nodes are used to represent the constraint information that guides the reservoir scheduling.

[0018] In this cascade reservoir group scheduling scenario, the specific reservoir entity nodes are Reservoir A, Reservoir B, and Reservoir C. These three nodes correspond to three actual reservoirs, and their physical state information includes key data such as water level and storage capacity. Runoff input nodes include runoff from River A into Reservoir A, runoff from River B into Reservoir B, and runoff from rainfall within the region into Reservoir C. These nodes reflect the natural water inflow from different sources. Water demand nodes include domestic water demand in urban residential areas, industrial water demand in industrial areas, and irrigation water demand in agricultural areas, representing the water supply needs of different users. Scheduling rule nodes include flood control scheduling rules, beneficial water use scheduling rules, and ecological flow guarantee rules. These rules provide constraints and guidance for reservoir scheduling.

[0019] Step S112: Extract the core attribute parameters of each node. For reservoir entity nodes, the core attribute parameters include current water level, effective storage capacity, inflow, outflow, and flood discharge capacity. For runoff input nodes, the core attribute parameters include runoff source type, runoff process morphology, and runoff forecast period length. For water demand nodes, the core attribute parameters include demand type, demand time period distribution, and water demand scale. For scheduling rule nodes, the core attribute parameters include rule application conditions, constraint index type, and index control range.

[0020] In this embodiment, for Reservoir A, the current water level refers to the elevation of the water surface at a certain moment; the effective storage capacity refers to the volume of water that Reservoir A can use for regulation during normal operation; the inflow rate is the amount of water flowing into Reservoir A per unit time; the outflow rate is the amount of water flowing out of Reservoir A per unit time; and the flood discharge capacity is the maximum flow rate that Reservoir A can safely discharge at a specific water level. For the runoff input node of River A flowing into Reservoir A, the runoff source type is surface runoff, and the runoff process exhibits seasonal variation characteristics, with larger runoff during the rainy season and smaller runoff during the dry season. The runoff forecast period refers to the length of time during which the runoff change can be predicted in advance. For the water demand node of the urban residential area, the demand type is domestic water use, and the demand period is concentrated in the morning, noon, and evening of the day. The water demand scale refers to the amount of water required by the urban residential area per unit time. For flood control scheduling rule nodes, the rule applies when the reservoir water level exceeds the flood control limit level. The constraint index types include flood discharge flow and water level control. The control range of the index clearly defines the upper and lower limits of flood discharge flow and the control range of water level.

[0021] Step S113: Establish directed edge connections between nodes. The reservoir entity node and the runoff input node form an input association through directed edges, and the reservoir entity node and the water demand node form an output association through directed edges. The scheduling rule node forms constraint associations with the reservoir entity node, the runoff input node, and the water demand node through directed edges respectively.

[0022] In this scenario, the runoff input node of river A flowing into reservoir A is connected to reservoir A via directed edges, forming an input relationship. This indicates that the runoff from river A flows into reservoir A, affecting the water volume of reservoir A. Reservoir A is connected to the water demand node of the urban residential area via directed edges, forming an output relationship. This means that reservoir A supplies water to the urban residential area. The flood control scheduling rule node is connected to reservoir A, the runoff input node of river A flowing into reservoir A, and the water demand node of the urban residential area via directed edges, forming a constraint relationship. This means that the flood control scheduling rule will constrain the operation of reservoir A, the runoff utilization of river A, and the water use of the urban residential area. Similarly, other reservoir entity nodes also establish similar directed edge connections with their corresponding runoff input nodes and water demand nodes, and the scheduling rule node also forms constraint relationships with other related nodes.

[0023] Step S114: Mark the initial path influence weight for each influence transmission path. The initial path influence weight is determined based on the actual degree of influence between nodes. The path influence weight of the runoff input node on the reservoir entity node is determined based on the correlation analysis of historical runoff and reservoir inflow. The path influence weight of the water demand node on the reservoir entity node is determined based on the priority of water demand guarantee. The path influence weight of the scheduling rule node on other nodes is determined based on the degree of enforcement of the rule.

[0024] For the influence transmission path from the runoff input node of River A into Reservoir A to Reservoir A, the runoff data of River A and the inflow data of Reservoir A over many years are analyzed. By calculating the correlation between the two, the initial influence weight of this path is determined. If historical data shows a high correlation between the two, the initial path influence weight is larger. For the influence transmission path from the water demand node of urban residential area to Reservoir A, the initial influence weight of this path is relatively large because the guarantee of domestic water supply has a high priority. The initial influence weight of the path from the water demand node of industrial area and agricultural area to the corresponding reservoir entity node is determined according to their priority. For the influence transmission path from the flood control scheduling rule node to Reservoir A, the initial path influence weight is large because flood control rules are highly mandatory. In contrast, the initial path influence weight from the beneficial scheduling rule node to related nodes is slightly lower according to its mandatory nature.

[0025] Step S120: Perform dynamic evolution update on the reservoir scheduling entity association graph based on hydrological time series data, update the real-time state attributes of each node through the node state propagation algorithm, adjust the path influence weight of the influence transmission path according to the real-time state attributes, and generate a time-varying scheduling association graph.

[0026] In this embodiment, after constructing the reservoir scheduling entity association graph, it needs to be dynamically updated based on real-time hydrological time-series data. Hydrological time-series data changes continuously over time, affecting the state of each node and the relationships between them. A node state propagation algorithm can update the real-time state attributes of nodes in a timely manner and adjust the path influence weights accordingly, thereby generating a time-varying scheduling association graph that reflects the real-time situation.

[0027] Step S121: Collect hydrological time-series data sequences, which include real-time monitoring data of reservoir entity nodes, forecast data of runoff input nodes, real-time feedback data of water demand nodes, and update data of scheduling rule nodes.

[0028] In this cascade reservoir group scheduling scenario, real-time monitoring data for reservoir entities is collected through sensors installed in the reservoirs, including hourly water level data, hourly inflow and outflow data for Reservoir A, etc. Forecast data for runoff input nodes is provided by a hydrological forecasting system, such as hourly forecasted flow data for the runoff from River A into Reservoir A over the next three days. Real-time feedback data for water demand nodes is collected by metering devices in each water-using area, such as hourly actual water consumption data for urban residential areas; comparing this with demand data yields feedback information. Updated data for scheduling rules nodes comes from the management department. When special weather conditions or policy adjustments occur, flood control scheduling rules may be supplemented or modified, and the aforementioned updated data will be collected promptly.

[0029] Step S122: Initialize the propagation parameters of the node state propagation algorithm. The propagation parameters include the propagation step size, the attenuation coefficient, and the convergence threshold. The propagation step size is determined according to the decision cycle of reservoir scheduling, and the attenuation coefficient is used to control the attenuation of the influence intensity of remote nodes.

[0030] The propagation step size is related to the decision-making cycle of reservoir scheduling. If the scheduling decision-making cycle of the cascade reservoir group is one day, then the propagation step size is set to one day, meaning that the node state is updated once a day. The attenuation coefficient must consider the influence of remote nodes. The influence of distant nodes on the target node decreases with increasing distance. For example, setting the attenuation coefficient to a value less than 1 gradually weakens the influence of remote nodes. The convergence threshold is a small value. When the change in node state is less than this threshold, it indicates that the node state tends to stabilize, and the propagation process can stop.

[0031] Step S123: Input the initial core attribute parameters of each node into the node state propagation algorithm. During the propagation process, each node receives the state influence value transmitted by its predecessor node through the influence propagation path. The state influence value is the product of the core attribute parameters of the predecessor node and the path influence weight of the corresponding influence propagation path.

[0032] The initial core attribute parameters of Reservoir A, such as its current water level and effective capacity, as well as the initial core attribute parameters of the runoff input node from River A flowing into Reservoir A, such as its initial runoff source type and runoff process morphology, and the initial core attribute parameters of other nodes, are input into the node state propagation algorithm. During propagation, Reservoir A receives state influence values ​​transmitted from the runoff input node from River A flowing into Reservoir A. These state influence values ​​are the product of the runoff flow rate of River A (a core attribute parameter) and the influence weight of the path between the two. Simultaneously, Reservoir A also receives similar state influence values ​​transmitted from other predecessor nodes.

[0033] Step S124: Accumulate all state influence values ​​received by each node, calculate the state update increment of the node in combination with the decay coefficient, and stop propagation when the state update increment of K consecutive iterations is less than the convergence threshold.

[0034] After receiving state influence values ​​from multiple predecessor nodes, including the runoff input node and related scheduling rule nodes from River A, Reservoir A accumulates these values. Then, it processes the accumulated results using a decay coefficient to calculate the state update increment of Reservoir A. This increment reflects the degree of change in the node's state. During the iteration process, the state update increment is continuously calculated. When the increment calculated for several consecutive iterations (e.g., 5 times) is less than the convergence threshold, it indicates that the node's state has basically stabilized, and the propagation process stops.

[0035] Step S125: Recalculate the path influence weights of each influence transmission path based on the updated real-time status attributes. For the path from the runoff input node to the reservoir entity node, adjust the path influence weights according to the current runoff forecast accuracy. For the path from the water demand node to the reservoir entity node, adjust the path influence weights according to the current water demand urgency. For the path from the scheduling rule node to other nodes, adjust the path influence weights according to the current applicability of the rule.

[0036] After the real-time status attributes of Reservoir A are updated, the weights of each influence transmission path are recalculated. If the deviation between the predicted runoff data of River A flowing into Reservoir A and the actual measured data is small, i.e., the current runoff prediction accuracy is high, the influence weight of the path from the runoff input node to Reservoir A is appropriately increased; conversely, if the prediction accuracy is low, the weight is appropriately decreased. When there is a water shortage in the urban residential area and the water demand is urgent, the influence weight of the path from the water demand node of the urban residential area to Reservoir A will increase. For the path from the flood control scheduling rule node to Reservoir A, the applicability of the rule is high during the flood season, and the path influence weight is large; while during the non-flood season, the applicability decreases, and the weight decreases accordingly.

[0037] Step S126: Apply the updated real-time status attributes and the adjusted path influence weights to the reservoir scheduling entity association graph to generate a time-varying scheduling association graph that changes dynamically over time.

[0038] The updated real-time status attributes of all reservoir entity nodes, including Reservoir A, Reservoir B, and Reservoir C, along with the adjusted path influence weights for each influence transmission path, replace the original attributes and weights in the reservoir scheduling entity association graph. After this update, the originally fixed association graph becomes a time-varying scheduling association graph that can dynamically change over time. This time-varying scheduling association graph can reflect the changes in the status of each node and the influence relationships between nodes in real time.

[0039] Step S130: Perform multi-agent collaborative reasoning on the time-varying scheduling association graph. Use the reservoir scheduling reasoning model to perform deep reasoning on the entity association relationships in the time-varying scheduling association graph to obtain the influence degree matrix and scheduling priority sequence between entities.

[0040] After generating the time-varying scheduling correlation graph, multi-agent collaborative reasoning is required. Multi-agent collaborative reasoning means comprehensively considering the interrelationships and influences among multiple entities such as reservoirs, runoff input, water demand, and scheduling rules. Through a specially constructed reservoir scheduling reasoning model, the correlations between entities are analyzed in depth to derive the influence degree matrix and scheduling priority sequence between entities, providing a basis for the subsequent formulation of scheduling schemes.

[0041] Step S131: Construct the reasoning framework of the reservoir scheduling reasoning model. The reasoning framework includes an entity association extraction layer, a relation reasoning layer, and a priority ranking layer. The entity association extraction layer is used to extract the direct association features between entities in the time-varying scheduling association graph. The relation reasoning layer is used to reason about the indirect influence relationship between entities. The priority ranking layer is used to determine the scheduling processing order of each node.

[0042] In the reasoning framework of the reservoir scheduling reasoning model, the entity association extraction layer extracts direct association features between entities from the time-varying scheduling association graph through specific algorithms and mechanisms. For example, the direct association features between reservoir A and the runoff input node of river A flowing into reservoir A. The relational reasoning layer further explores indirect influence relationships between entities based on these direct association features. For instance, the runoff of river A affects reservoir A, which in turn affects the water volume of reservoir B; these indirect relationships are derived by the relational reasoning layer. The priority ranking layer determines the order in which nodes are processed during scheduling based on factors such as their importance and scope of influence.

[0043] Step S132: Input the real-time state attribute matrix of the nodes and the path influence weight adjacency matrix of the time-varying scheduling association graph into the entity association extraction layer. Calculate the direct association strength between nodes through the graph attention mechanism of the entity association extraction layer to generate a direct association feature matrix. The elements in the direct association feature matrix represent the degree of direct influence between two corresponding nodes.

[0044] The real-time state attributes of each node in the time-varying scheduling association graph are organized into a matrix form, namely the node real-time state attribute matrix. Simultaneously, the influence weights of each path are also organized into an adjacency matrix form, namely the path influence weight adjacency matrix. After inputting these two matrices into the entity association extraction layer, the graph attention mechanism focuses on the important associations between nodes and calculates the direct association strength between nodes. For example, by analyzing the association between Reservoir A and the water demand node in the urban residential area through the graph attention mechanism, their direct association strength is obtained. The direct association strengths between all nodes are then integrated to generate a direct association feature matrix, where each element corresponds to the degree of direct influence between two nodes.

[0045] Step S1321: Perform a linear transformation on the node real-time state attribute matrix to generate a query feature matrix, a key feature matrix, and a value feature matrix. The query feature matrix and the key feature matrix are used to calculate the attention score, and the value feature matrix is ​​used to generate association features.

[0046] A linear transformation is performed on the real-time state attribute matrix of the nodes, generating query feature matrix, key feature matrix, and value feature matrix using different transformation parameters. The query feature matrix represents the features of each node as the query party, the key feature matrix represents the features of each node as the query target, and the two interact to calculate the attention score. The value feature matrix is ​​the foundation matrix used to generate the final associated features. For example, after the linear transformation, the real-time state attributes of Reservoir A have corresponding feature vectors in the query feature matrix, key feature matrix, and value feature matrix.

[0047] Step S1322: For each node in the time-varying scheduling association graph, the original attention score of that node to all other nodes is obtained by performing a dot product operation between the corresponding row vector of the query feature matrix and all column vectors of the key feature matrix.

[0048] Taking Reservoir A as an example, the row vector corresponding to Reservoir A is found in the query feature matrix. Then, the row vector is multiplied by the column vectors corresponding to all nodes in the key feature matrix. The result of the dot product operation is the original attention score of Reservoir A to each other node. These scores reflect the initial degree of attention association between Reservoir A and other nodes.

[0049] Step S1323: The original attention score is indexed and the attention weight is obtained by row normalization. The attention weight reflects the importance of the direct relationship between nodes.

[0050] The raw attention scores are then indexed to ensure they fall within a suitable range, avoiding values ​​that are too high or too low. Next, the indexed scores are normalized row-wise: for each node's corresponding row score, each score in that row is divided by the sum of all scores in that row to obtain the attention weights. These attention weights better reflect the importance of the direct relationships between nodes; higher weights indicate more important relationships.

[0051] Step S1324: Perform a weighted summation operation on the attention weights and value feature matrices to generate the associated feature vector of each node. Arrange the associated feature vectors of all nodes in rows to generate a direct association feature matrix. The number of rows and columns of the direct association feature matrix are equal to the total number of nodes.

[0052] The calculated attention weights are then summed with the corresponding elements in the value feature matrix. In other words, the association feature vector of each node is obtained by summing the value feature vectors of other nodes according to their attention weights. For example, the association feature vector of Reservoir A is obtained by multiplying the value feature vectors of other nodes by their respective attention weights and then summing them. The association feature vectors of all nodes are arranged row-wise to generate a direct association feature matrix. This matrix has the same number of rows and columns as the total number of nodes, fully reflecting the direct association features between nodes.

[0053] Step S133: Input the direct association feature matrix into the relationship reasoning layer, mine the effective indirect influence paths between entities through the path reasoning algorithm of the relationship reasoning layer, and generate the indirect association feature matrix based on the effective indirect influence paths.

[0054] After inputting the direct association feature matrix into the relational reasoning layer, the path reasoning algorithm analyzes the paths between entities and identifies indirect paths that can effectively transmit influence. For example, reservoir A indirectly affects the water allocation of the industrial zone by supplying water to the urban residential area; if such indirect paths are effective, they will be identified. Based on these effective indirect influence paths, the degree of indirect influence between entities is calculated, thereby generating an indirect association feature matrix.

[0055] Step S1331: Initialize the path length range and weight threshold of the path reasoning algorithm.

[0056] The path length range refers to the upper limit of the length of indirect paths considered by the algorithm. For example, setting the path length range to 3 means that only indirect paths that pass through two intermediate nodes are considered. The weight threshold is the standard for judging whether an indirect path is valid. When the cumulative influence weight of a path reaches or exceeds the threshold, the path is considered a valid indirect influence path.

[0057] Step S1332: For each node pair in the time-varying scheduling association graph, extract all directed paths whose lengths are within the path length range. Each directed path consists of continuous influence propagation paths.

[0058] Taking the node pair of Reservoir A and the industrial zone water demand node as an example, within the path length range, extract all directed paths from Reservoir A to the industrial zone water demand node. These paths may be Reservoir A - Urban Residential Area Water Demand Node - Industrial Zone Water Demand Node, or Reservoir A - Reservoir B - Industrial Zone Water Demand Node, etc. Each path consists of continuous influence transmission paths.

[0059] Step S1333: Calculate the cumulative influence weight of each directed path, whereby the cumulative influence weight is the product of the path influence weights of all influence propagation paths in the directed path.

[0060] For each extracted directed path, the path influence weights of each influence transmission path in the path are multiplied together to obtain the cumulative influence weight of the directed path. For example, for the directed path Reservoir A - Urban Residential Area Water Demand Node - Industrial Area Water Demand Node, assuming the path influence weight from Reservoir A to the Urban Residential Area Water Demand Node is W1, and the path influence weight from the Urban Residential Area Water Demand Node to the Industrial Area Water Demand Node is W2, then the cumulative influence weight of this directed path is the product of W1 and W2.

[0061] Step S1334: Mark the directed paths with cumulative influence weights greater than the weight threshold as effective indirect influence paths, and record the starting node, ending node, and cumulative influence weight of the effective indirect influence paths.

[0062] After calculating the cumulative impact weights of all directed paths, they are compared with a preset weight threshold. If the cumulative impact weight of a directed path is greater than or equal to the weight threshold, it is marked as a valid indirect impact path. Simultaneously, the starting node, ending node, and corresponding cumulative impact weight of this valid indirect impact path are recorded in detail. For example, if the cumulative impact weight of the path Reservoir A - Urban Residential Area Water Demand Node - Industrial Area Water Demand Node is greater than the weight threshold, it is marked as a valid indirect impact path, and the starting node is Reservoir A, the ending node is the Industrial Area Water Demand Node, and the corresponding cumulative impact weight is recorded.

[0063] Step S1335: For each node pair, summarize the cumulative influence weights of all effective indirect influence paths, use the summary result as the indirect influence strength of the node pair, and generate an indirect association feature matrix based on the indirect influence strengths of all node pairs. The number of rows and columns of the indirect association feature matrix are equal to the total number of nodes.

[0064] For each pair of nodes, the cumulative influence weights of all valid indirect influence paths belonging to that pair are summed. This summation result represents the indirect influence strength of that pair. For example, for the pair of nodes Reservoir A and the industrial zone's water demand, if multiple valid indirect influence paths exist, the cumulative influence weights of these paths are added together to obtain the indirect influence strength between them. The indirect influence strengths of all pair of nodes are arranged in the order of their corresponding nodes to form an indirect association feature matrix. The number of rows and columns in this matrix are consistent with the total number of nodes, and each element in the matrix represents the indirect influence strength between the corresponding two nodes.

[0065] Step S134: Merge the direct association feature matrix and the indirect association feature matrix to generate a comprehensive association feature matrix. Input the comprehensive association feature matrix into the influence degree calculation module of the relationship reasoning layer. Decompose the comprehensive association feature matrix into entity influence vectors through a matrix decomposition algorithm. Calculate the influence degree matrix between entities based on the entity influence vectors. The row vectors in the influence degree matrix represent the comprehensive influence degree of the corresponding node on all other nodes.

[0066] The direct and indirect association feature matrices are fused. The fusion method can be based on a set weight ratio, determined by the actual importance of the association features. The fused matrix generates a comprehensive association feature matrix. This comprehensive association feature matrix is ​​then input into the influence calculation module. The matrix factorization algorithm processes it, decomposing it into influence vectors for each entity. These entity influence vectors reflect the entity's own influence characteristics. Based on these entity influence vectors, the interaction relationships between the vectors are calculated to obtain the comprehensive influence degree between entities, thus forming an influence degree matrix. Each row of vectors corresponds to the comprehensive influence degree of a node on all other nodes.

[0067] Step S135: Input the influence matrix into the priority ranking layer, and use the multi-factor decision-making algorithm of the priority ranking layer to give a comprehensive score to each node. The multi-factor decision-making algorithm combines the influence range, influence intensity and current urgency of the node, and sorts the nodes from high to low according to the comprehensive score to generate a scheduling priority sequence.

[0068] After inputting the influence matrix into the priority ranking layer, the multi-factor decision-making algorithm first analyzes the influence range of each node, i.e., the number of other nodes that the node can influence; then, it evaluates the strength of the node's influence by referring to the row vector value corresponding to that node in the influence matrix; simultaneously, it considers the urgency of the node's current state, such as whether the reservoir water level is close to the warning level or whether the water demand is urgent. Each node is scored based on these three factors, with different weights assigned according to the importance of each factor during the scoring process. Finally, all nodes are ranked according to their comprehensive scores, with nodes scoring higher ranked first, forming a scheduling priority sequence. This scheduling priority sequence clearly defines the order in which nodes are processed during the scheduling process.

[0069] Step S140: Establish a joint reservoir scheduling scheme based on the influence matrix and scheduling priority sequence. The joint reservoir scheduling scheme includes a water source allocation sub-scheme, a demand response sub-scheme, and a rule coordination sub-scheme.

[0070] After obtaining the influence matrix and scheduling priority sequence, a joint reservoir scheduling scheme is constructed based on these. The influence matrix provides a basis for the degree of influence of various resource allocations and rule coordination, while the scheduling priority sequence determines the order of processing. Combining these two, three sub-schemes are formulated for water resource allocation, demand response, and rule coordination, and then integrated into a complete joint reservoir scheduling scheme.

[0071] Step S141: Analyze the row vectors of reservoir entity nodes in the influence matrix, extract the influence value of each reservoir entity node on the runoff input node, and determine the water source contribution weight of each runoff input node by combining the order of runoff input nodes in the scheduling priority sequence. The water source contribution weight is positively correlated with the influence value.

[0072] The row vectors corresponding to the reservoir entity nodes in the influence matrix are analyzed to extract the influence value of each reservoir entity node on each runoff input node. For example, the influence values ​​of Reservoir A on the runoff input node of River A flowing into Reservoir A, and the runoff input node of River B flowing into Reservoir B are extracted from the row vector corresponding to Reservoir A. Simultaneously, the order of runoff input nodes in the scheduling priority sequence is examined, and the influence values ​​are combined with this order to determine the water source contribution weight of each runoff input node. The larger the influence value, the larger the corresponding water source contribution weight, meaning that the contribution of that runoff input node to water supply is assigned a higher weight.

[0073] Step S142: Establish a water source allocation sub-scheme based on the water source contribution weight. The water source allocation sub-scheme includes the water inflow allocation ratio and time period allocation plan of each runoff input node to the reservoir entity node.

[0074] Based on the determined water contribution weights of each runoff input node, the proportion of water allocated to the corresponding reservoir node by each runoff input node is calculated. For example, if the water contribution weight of the runoff input node flowing from river A into reservoir A is high, then its proportion of water allocated to reservoir A will be correspondingly larger. Simultaneously, considering the seasonal variations in runoff and the operational needs of the reservoir, a time-period allocation plan is developed, specifying the proportion and volume of water supplied by each runoff input node to the reservoir node during different time periods. The water allocation proportions and time-period allocation plans are then integrated to form a water resource allocation sub-scheme. This sub-scheme ensures that the reservoir can rationally receive and utilize the water from each runoff input.

[0075] Step S143: Analyze the column vectors of water demand nodes in the influence matrix, extract the influence values ​​of each reservoir entity node on the water demand nodes, and construct the water demand satisfaction matrix by combining the order of water demand nodes in the scheduling priority sequence. The elements in the water demand satisfaction matrix represent the water supply guarantee capacity of the corresponding reservoir entity node for the water demand nodes.

[0076] The column vectors corresponding to water demand nodes in the influence matrix are analyzed to obtain the influence value of each reservoir entity node on each water demand node. For example, from the column vectors corresponding to water demand nodes in urban residential areas, the influence values ​​of Reservoir A, Reservoir B, and Reservoir C are extracted. Referring to the order of water demand nodes in the scheduling priority sequence, these influence values ​​are organized to construct a water demand satisfaction matrix. Each element in the matrix represents the water supply guarantee capacity of a certain reservoir entity node for a certain water demand node; the higher the value, the stronger the water supply guarantee capacity of that reservoir for that water demand node.

[0077] Step S144: Establish a demand response sub-scheme based on the water demand satisfaction matrix. The demand response sub-scheme includes the allocation of water supply sources, water supply time period arrangement and water supply flow adjustment plan for each water demand node.

[0078] Based on the water demand satisfaction matrix, the water supply source is determined for each water demand node, specifying which reservoir nodes will supply water and their respective supply ratios. According to the demand period distribution of the water demand nodes and the water supply capacity of the reservoir nodes, the water supply period is rationally scheduled to match peak water consumption periods. Simultaneously, based on the water demand scale and the availability of water supply sources, a water supply flow regulation plan is developed to ensure that the water supply flow can meet demand and remain stable. Integrating these three parts forms a demand response sub-scheme, which can effectively address the water supply issues of different water demand nodes.

[0079] Step S1441: Perform row normalization on the water demand satisfaction matrix to obtain the water supply allocation ratio of each reservoir entity node to the water demand node. The water supply allocation ratio represents the water supply contribution share of the reservoir entity node to the water demand node.

[0080] A row-level normalization operation is performed on the water demand satisfaction matrix. This involves taking each row (representing a reservoir entity node) and dividing each element of that row (representing the water supply guarantee capacity for a specific water demand node) by the sum of all elements in that row. The result is the water supply allocation ratio of that reservoir entity node to the corresponding water demand node. This allocation ratio reflects the contribution share of that reservoir entity node in satisfying the water demand of its corresponding water demand node. For example, after normalization, the row corresponding to Reservoir A yields its water supply allocation ratios to urban residential areas, industrial areas, and agricultural areas, representing the share of water supplied by Reservoir A to these three water demand nodes, respectively.

[0081] Step S1442: According to the order of water demand nodes in the scheduling priority sequence, allocate water supply sources to each water demand node in turn.

[0082] According to the priority sequence of water demand nodes, the nodes with the highest priority are processed first. For the top-ranked water demand node, the main water supply source is allocated from reservoir entities with strong water supply security, based on the water supply allocation ratio of each reservoir entity node, while other reservoir entities serve as supplementary sources. After allocating water supply sources for this node, the next priority water demand node is allocated water supply sources in the same way, ensuring that each node receives adequate water supply security and prioritizing the needs of high-priority nodes.

[0083] Step S1443: Analyze the water demand period distribution characteristics of water demand nodes, and combine the inflow process and outflow capacity constraints of reservoir physical nodes to establish a water supply period arrangement for each water demand node. The water supply period arrangement matches the peak water demand period of the water demand node with the peak water supply capacity period of the reservoir physical node.

[0084] A thorough analysis of the water demand distribution at each water demand node is conducted to pinpoint the specific time periods of peak water consumption. Simultaneously, the inflow process of each reservoir is studied to understand the inflow situation at different times and the reservoir's outflow capacity constraints—that is, the maximum amount of water the reservoir can release at different times. Based on this information, a water supply schedule is developed for each water demand node, aiming to align the peak water demand periods of each node with the peak water supply capacity periods of the reservoirs supplying it, thus ensuring sufficient water supply during peak demand periods.

[0085] Step S1444: Based on the water demand flow scale and water supply allocation ratio of the water demand nodes, calculate the initial water supply flow of each reservoir entity node to the water demand nodes, and adjust the initial water supply flow in combination with the current water level and reservoir capacity status of the reservoir entity nodes to generate a water supply flow regulation plan.

[0086] First, based on the water demand scale of each water demand node and the corresponding water supply allocation ratio of each reservoir entity node, the initial water supply flow of each reservoir entity node to the corresponding water demand node is obtained through multiplication. Then, the current water level and storage capacity of each reservoir entity node are checked. If the current water level is high and the storage capacity is sufficient, the initial water supply flow may be appropriately increased; if the water level is low and the storage capacity is limited, the initial water supply flow will be appropriately decreased. After the above adjustments, a water supply flow regulation plan is formed, which clarifies the water supply flow of each reservoir entity node to the corresponding water demand node at different times.

[0087] Step S1445: Integrate water supply source allocation, water supply time period arrangement and water supply flow adjustment plan to generate a complete demand response sub-scheme. The demand response sub-scheme also includes the setting of flow monitoring points and deviation adjustment mechanism during the water supply process. When the deviation between the actual water supply flow and the planned water supply flow exceeds the preset range, the deviation adjustment mechanism is activated to dynamically correct the water supply flow.

[0088] The initial demand response sub-scheme is formed by integrating the allocation of water supply sources, the scheduling of water supply periods, and the water supply flow regulation plan. Based on this, additional flow monitoring points are set up at key locations along the water transmission line to monitor the actual water supply flow in real time. Simultaneously, a deviation adjustment mechanism is established, setting a deviation range. When the difference between the actual and planned water supply flow exceeds this range, the deviation adjustment mechanism is immediately activated, dynamically correcting the water supply flow by adjusting reservoir gate openings and other methods to ensure that the water supply flow remains within the planned range, ultimately forming a complete demand response sub-scheme.

[0089] Step S145: Analyze the relevant row vectors of the scheduling rule nodes in the influence matrix, extract the influence values ​​of the scheduling rule nodes on other types of nodes, and determine the constraint strength coefficient of each scheduling rule by combining the order of the scheduling rule nodes in the scheduling priority sequence. The constraint strength coefficient is positively correlated with the influence value.

[0090] Locate the row vectors corresponding to the scheduling rule nodes in the influence matrix, and extract the influence values ​​of each scheduling rule node on other types of nodes such as reservoir entity nodes, runoff input nodes, and water demand nodes. Combined with the arrangement order of the scheduling rule nodes in the scheduling priority sequence, analyze and process these influence values ​​to determine the constraint strength coefficient of each scheduling rule. The magnitude of the constraint strength coefficient is positively correlated with the influence value; that is, the larger the influence value, the larger the corresponding constraint strength coefficient, indicating that the scheduling rule has a stronger constraint effect on other nodes.

[0091] Step S146: Establish a rule coordination sub-scheme based on the constraint strength coefficient. The rule coordination sub-scheme includes the applicable priority ranking of scheduling rules, the coordination strategy of conflicting rules, and the dynamic adjustment range of rule indicators.

[0092] Based on the constraint strength coefficients of each scheduling rule, the rules are prioritized for application, with rules having higher constraint strength coefficients listed first and applied preferentially. When conflicts arise between different scheduling rules, they are handled according to the pre-defined coordination strategy in the rule coordination sub-scheme. For example, when flood control scheduling rules and beneficial utilization scheduling rules conflict on a certain decision, the coordination strategy determines which rule takes precedence. Simultaneously, the dynamic adjustment range of each rule's indicators is clearly defined, allowing adjustments to these indicators within this range during actual scheduling. Integrating these three aspects forms the rule coordination sub-scheme, ensuring the orderly and effective application of scheduling rules in practice.

[0093] Step S147: Integrate the water source allocation sub-scheme, demand response sub-scheme, and rule coordination sub-scheme to generate a joint reservoir scheduling scheme that includes multi-dimensional scheduling objectives.

[0094] The water resource allocation sub-scheme, demand response sub-scheme, and rule coordination sub-scheme are integrated. During the integration process, smooth coordination between the sub-schemes is ensured, and no contradictions or conflicts are found. The integrated reservoir joint operation scheme covers multiple dimensions of operation objectives, including ensuring water supply security, meeting flood control requirements, improving water resource utilization efficiency, and maintaining ecological balance. This integrated reservoir joint operation scheme comprehensively considers water resource allocation, demand response, and rule coordination.

[0095] Step S150: Generate a cascade scheduling instruction set based on the reservoir joint scheduling scheme, and send the cascade scheduling instruction set to the reservoir control center to control the regulation of reservoir gates, the adjustment of unit output and the distribution of water flow.

[0096] Guided by the joint reservoir operation plan, the various components of the plan are translated into specific cascade operation instructions. These instructions cover multiple aspects, including the opening and closing of reservoir gates, adjustment of generator output, and allocation of water flow. All these instructions are compiled into a cascade operation instruction set and sent to the reservoir control center. The control center then uses this instruction set to control the relevant equipment in each reservoir, ensuring the reservoir group operates according to the joint operation plan.

[0097] Step S151: Analyze the water source allocation sub-scheme in the joint reservoir scheduling scheme, extract the inflow forecast value and time period allocation plan of each reservoir entity node, and calculate the target water storage level for each time period by combining the current water level and reservoir capacity curve of the reservoir entity node.

[0098] The water allocation sub-scheme in the joint reservoir operation plan is analyzed to obtain the predicted inflow and time-sharing allocation plan for each reservoir entity node at different time periods. Simultaneously, current water level data and reservoir capacity curves are collected for each reservoir entity node; the capacity curves reflect the relationship between reservoir water level and capacity. Based on the predicted inflow, time-sharing allocation plan, current water level, and capacity curves, the target water storage level that each reservoir entity node should reach in each time period is determined through methods such as water balance calculations. The determination of the target water storage level must ensure that the reservoir can both meet subsequent water demand and guarantee reservoir safety.

[0099] Step S152: Generate reservoir gate regulation instructions based on the target water level. The reservoir gate regulation instructions include gate type, regulation period, target opening degree and opening change rate. For flood discharge gates, the number of gate opening groups and opening degree are determined based on the predicted inflow and flood control limit water level.

[0100] Based on the calculated target water levels for each time period, reservoir gate regulation instructions are generated. The instructions specify the type of gate to be regulated, such as flood discharge gates and power generation gates; specify the specific time periods for gate regulation; set the target gate opening degree; and determine the rate of change of gate opening degree to ensure a smooth regulation process. For flood discharge gates, a comprehensive analysis is conducted combining the predicted inflow and the flood control limit level. When a large inflow may cause the water level to exceed the flood control limit level, the number of gate groups to be opened and the opening degree of each group are determined to ensure timely flood discharge and safeguard the reservoir's flood control safety.

[0101] For example, step S1521: determine the type and number of reservoir gates. The gate types include flood discharge gates, power generation gates, irrigation gates and water conveyance gates. Each type of gate corresponds to different functions and operating characteristics.

[0102] A comprehensive review of the sluice gates equipped in the reservoir was conducted to clarify their types and quantities. Flood discharge gates are primarily used for releasing floodwaters during floods to ensure reservoir safety; power generation gates work in conjunction with generator sets to control the outflow and drive the generators to produce electricity; irrigation gates are used to supply water to agricultural irrigation areas; and water conveyance gates are used to deliver domestic and industrial water to cities and industrial zones. Each type of gate differs in its structure, operation method, and control precision. For example, flood discharge gates are typically large and require rapid opening and closing capabilities; while power generation gates require high precision in controlling their opening to ensure stable power output.

[0103] Step S1522: Based on the difference between the target water level and the current water level, determine the type of gate to be adjusted. When the target water level is higher than the current water level, prioritize adjusting the inlet gate to increase the inflow. When the target water level is lower than the current water level, prioritize adjusting the flood discharge gate or the outlet gate to increase the outflow.

[0104] Calculate the difference between the target water level and the current water level, and determine the type of gate to be adjusted based on the sign of the difference. If the target water level is higher than the current water level, it indicates that the reservoir's storage capacity needs to be increased. In this case, prioritize adjusting the inlet gate, increasing the opening of the inlet gate to increase the inflow and gradually raise the reservoir water level to the target water level. If the target water level is lower than the current water level, it indicates that the reservoir's storage capacity needs to be reduced. In this case, prioritize adjusting the flood discharge gate or outlet gate, increasing the opening of these gates to increase the outflow and gradually lower the reservoir water level to the target water level.

[0105] Step S1523: Calculate the target outflow for each time period based on the predicted inflow and the target water level. Determine the target opening degree of each gate according to the target outflow and the gate flow characteristic curve. The gate flow characteristic curve represents the relationship between the gate opening degree and the flow through the gate.

[0106] By combining the predicted inflow and target water level, the target outflow for each time period is calculated using the water balance principle to ensure that the reservoir water level reaches the target water level at the end of that period. Each gate has a corresponding gate flow characteristic curve, which is obtained through fitting multiple experiments and actual operating data, accurately reflecting the correspondence between gate opening and throughflow. For example, for the flood discharge gate of Reservoir A, the corresponding throughflow can be found from the gate's flow characteristic curve when the gate opening is a certain value. After determining the target outflow, the target opening of each gate that can achieve the target outflow is found by working backwards based on the function and flow characteristic curve of each gate. If the target outflow needs to be shared by multiple gates, the opening of each gate is reasonably allocated according to its flow characteristic curve, so that the sum of the throughflow of each gate equals the target outflow.

[0107] Step S1524: Calculate the gate opening change rate based on the length of the adjustment period and the range of the target opening change. The opening change rate must meet the mechanical performance limitations of the gate operation and the requirements for water flow stability.

[0108] The length of the adjustment period refers to the allowable time from the current gate opening to the target opening. The range of change in the target opening is the difference between the current opening and the target opening. Dividing the range of change in the target opening by the length of the adjustment period yields the gate opening change rate. During the calculation, the mechanical performance parameters of the gate must be considered to ensure that the calculated opening change rate does not exceed the maximum change rate that the gate's mechanical structure can withstand, preventing damage to gate components due to excessively rapid opening changes. Simultaneously, the requirements for water flow stability must be considered. If the opening change rate is too rapid, it may lead to drastic changes in water flow, causing water hammer and affecting the safety of the reservoir and downstream river channels. Therefore, the calculated opening change rate must be controlled within a range that ensures stable water flow.

[0109] Step S1525: Generate independent gate adjustment instructions for each type of gate. The gate adjustment instructions include a gate type identifier, adjustment start time, adjustment end time, target opening value, and maximum opening change rate. When multiple types of gates need to be adjusted simultaneously, determine the order of adjustment for each gate.

[0110] Independent gate regulation commands are generated for different types of gates, such as flood discharge gates, power generation gates, irrigation gates, and water conveyance gates. Each command includes an identifier to distinguish the gate type, such as "Flood Discharge Gate-1" and "Power Generation Gate-2". The start and end times of regulation are determined according to the regulation period to ensure that the gate opening adjustment is completed within the specified time. The target opening value is the final opening determined through calculation, and the maximum rate of change of opening is an upper limit set to ensure gate safety and stable water flow. When multiple types of gates need to be regulated simultaneously, the order of regulation is determined according to scheduling priority and the degree of impact on water flow. For example, during flood control, the regulation priority of flood discharge gates is higher than that of power generation and irrigation gates, and they should be regulated first; under the condition of ensuring water supply, the regulation of irrigation and water conveyance gates may take precedence over that of power generation gates.

[0111] Step S1526: Set safety verification conditions in the gate adjustment command. The safety verification conditions include the upper and lower limits of gate opening, the limit of the difference in opening between adjacent gates, and the limit of the acceleration of opening change. When the safety verification conditions are triggered during the gate adjustment process, the adjustment will be automatically stopped and an alarm message will be issued.

[0112] The upper and lower limits of the gate opening are determined based on the gate's design parameters and operational requirements. The upper limit is the maximum safe opening degree of the gate, and the lower limit is the minimum safe closing degree, preventing malfunctions caused by over-opening or over-closing. The limit on the difference in opening degree between adjacent gates is set for gates in the same group or adjacent gates to avoid uneven water flow distribution due to excessive differences in opening degrees between adjacent gates, which could affect the gate's structural safety and water flow stability. The limit on the acceleration of opening change is to further control changes in gate opening, preventing rapid changes in the rate of change within a short period and protecting the gate's drive mechanism. During gate adjustment, parameters such as the gate opening, the rate of change of opening degree, and the difference in opening degree between adjacent gates are monitored in real time. When these parameters reach the safety verification conditions, the gate adjustment operation is immediately and automatically stopped, and an alarm message is sent to the reservoir control center through the system, prompting staff to check and handle the situation promptly.

[0113] Step S153: Analyze the demand response sub-scheme in the joint reservoir scheduling scheme, extract the water supply flow regulation plan of each reservoir entity node, and calculate the target power generation output for each time period by combining the power generation head characteristic curve of the reservoir entity node.

[0114] The water supply flow regulation plans for each reservoir entity node at different time periods are obtained from the demand response sub-scheme, clarifying the flow rate that needs to be allocated to power generation in each time period. The power generation head characteristic curve of the reservoir entity node reflects the relationship between the reservoir water level and the power generation head. The corresponding power generation head can be obtained from this curve using the current reservoir water level. Based on the power generation water flow rate in the water supply flow regulation plan and the obtained power generation head, combined with the efficiency characteristics of the generator set, the target power generation output for each time period is calculated. For example, if the power generation water flow rate in the water supply flow regulation plan of Reservoir B is a certain value for a certain time period, the power generation head can be found from the power generation head characteristic curve based on the water level of Reservoir B at this time, and then combined with the efficiency parameters of the generator set, the target power generation output of Reservoir B for that time period can be calculated.

[0115] Step S154: Generate unit output adjustment instructions based on the target power generation output. The unit output adjustment instructions include the unit number, the number of units started, the target output, and the output adjustment rate. Coordinate the flow allocation of power generation water and other water use according to the priority relationship between the water supply flow regulation plan and the target power generation output.

[0116] Each generator set has a unique unit number. The number of units to be started is determined based on the target power output and the rated output of each unit. For example, if the target power output is a certain value, and the rated output of a single unit is another value, the number of units to be started can be calculated. The target output is the output value that each unit needs to achieve. The output adjustment rate refers to the speed at which the unit's output changes from its current value to the target output. This output adjustment rate must conform to the operating characteristics of the generator set to avoid affecting the stable operation of the unit due to excessively rapid adjustments. When generating instructions, if there is a flow conflict between power generation water and irrigation water, domestic water, or other water uses in the water supply flow regulation plan, coordination is carried out according to the preset priority relationship. When domestic water and irrigation water have a higher priority than power generation water, the flow of power generation water is appropriately reduced, and the flow of other water uses is increased. In this case, the target output and the number of units to be started in the unit output adjustment instruction are adjusted accordingly. When the priority of power generation water is higher, the flow of power generation water is guaranteed first, and then the flow of other water uses is allocated.

[0117] Step S155: Analyze the rule coordination sub-scheme in the joint reservoir scheduling scheme, extract the dynamic adjustment range of the rule indicators of each reservoir entity node, and calculate the target water conveyance flow for each time period by combining the hydraulic connection relationship of the inter-reservoir water conveyance system.

[0118] The dynamic adjustment range of rule indicators for each reservoir entity node is obtained from the rule coordination sub-scheme, such as the upper and lower limits of ecological flow and the fluctuation range of water conveyance flow. The hydraulic connection relationship of the inter-reservoir water conveyance system includes parameters such as the length of the water conveyance channel between each reservoir, the channel roughness, and the difference in water level between upstream and downstream. Based on these parameters and the dynamic adjustment range of rule indicators, combined with the water demand of the water demand nodes and the water supply plan for each time period, the target water conveyance flow for each time period is calculated. For example, when water is conveyed from Reservoir A to Reservoir B, based on the hydraulic connection relationship between the two reservoirs, considering the head loss and water conveyance efficiency during the water conveyance process, combined with the water demand of Reservoir B and the dynamic adjustment range of rule indicators, the target water conveyance flow from Reservoir A to Reservoir B for each time period is calculated.

[0119] Step S156: Generate a water conveyance flow allocation instruction based on the target water conveyance flow. The water conveyance flow allocation instruction includes the water conveyance channel number, target flow, regulation period, and operation plan of the control gates along the line. According to the joint operation requirements of the cascade reservoirs, coordinate the connection of water conveyance flow between the upstream and downstream reservoirs.

[0120] Water conveyance channel numbers are used to identify different water conveyance channels. For example, "Channel-AB" indicates a water conveyance channel connecting reservoir A and reservoir B. The target flow rate is the amount of water that needs to be conveyed through this channel at each time period. The regulation period corresponds to the calculation period of the target flow rate. The operation plan for the control gates along the route is formulated based on the target flow rate and the hydraulic characteristics of the water conveyance channel, specifying the operation mode and opening changes of each control gate to ensure that the water flow rate remains stable within the target flow rate range. When formulating instructions, the joint operation requirements of the cascade reservoirs are fully considered. The water flow rate of the upstream reservoir must match the receiving capacity of the downstream reservoir. For example, when upstream reservoir A conveys water to downstream reservoir B, the target flow rate in its water flow allocation instruction must be determined based on the current water level, capacity, and outflow plan of reservoir B to ensure that reservoir B can smoothly receive and process the conveyed water, avoiding situations where the water level in reservoir B is too high or too low.

[0121] Step S157: Integrate reservoir gate regulation instructions, unit output adjustment instructions, and water flow distribution instructions to generate a cascade scheduling instruction set with timestamp alignment. Each cascade scheduling instruction in the cascade scheduling instruction set includes the execution entity identifier, instruction effective time, target parameters, and operation constraints.

[0122] The generated reservoir gate regulation commands, generator output adjustment commands, and water flow distribution commands are integrated, and the timestamps of all commands are aligned to ensure that commands at the same point in time can be executed collaboratively. The executing entity identifier clarifies the target of the command, such as "Reservoir A Gate Control System" or "Reservoir B Generator Unit." The command effective time is the time when the command begins execution, consistent with the start time of the regulation period. Target parameters include target gate opening, target generator output, and target water flow. Operational constraints refer to the restrictions that must be followed during command execution, such as the gate opening not exceeding upper or lower limits and the generator output adjustment rate not exceeding specified values. Through integration, a complete cascade scheduling command set is formed, covering all operational commands during the operation of the cascade reservoir group.

[0123] Step S158: Send the cascade scheduling instruction set to the reservoir control center so that the reservoir control center can control the gate actuators, generator sets and water conveyance control equipment of each reservoir according to the cascade scheduling instruction set.

[0124] The cascade scheduling instruction set is transmitted to the reservoir control center via a dedicated communication network. Upon receiving the instruction set, the reservoir control center parses and verifies the instructions to confirm their completeness and validity. Subsequently, based on the executing entity identifier in the instruction, the corresponding instructions are distributed to the gate actuators, generator control systems, and water conveyance control equipment controllers of each reservoir. The gate actuators adjust the gate opening according to the gate adjustment instructions; the generator sets adjust their operating status according to the unit output adjustment instructions to achieve the target output; and the water conveyance control equipment operates the control gates along the route according to the water flow distribution instructions to control the water flow. During execution, the reservoir control center receives feedback information from each device in real time, monitors the execution of instructions, and ensures that all operations are accurately executed according to the requirements of the cascade scheduling instruction set.

[0125] Figure 2 The diagram illustrates exemplary hardware and software components of a reservoir scheduling and management system 100 that can implement the ideas of this application, according to some embodiments of this application. For example, a processor 120 can be used on the reservoir scheduling and management system 100 and to perform the functions in this application.

[0126] The reservoir scheduling and management system 100 can be a general-purpose server or a special-purpose server, both of which can be used to implement the reservoir scheduling and management method of this application. Although only one server is shown in this application, for convenience, the functions described in this application can be implemented in a distributed manner on multiple similar platforms to balance the load.

[0127] For example, the reservoir scheduling and management system 100 may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the reservoir scheduling and management system 100 may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The methods of this application can be implemented according to these program instructions. The reservoir scheduling and management system 100 also includes an I / O interface 150 between the computer and other input / output devices.

[0128] For ease of explanation, only one processor is described in the reservoir scheduling and management system 100. However, it should be noted that the reservoir scheduling and management system 100 of this application may also include multiple processors, and therefore the steps performed by one processor as described in this application may also be performed jointly or individually by multiple processors. For example, if the processor of the reservoir scheduling and management system 100 performs steps A and B, it should be understood that steps A and B may also be performed jointly by two different processors or individually by one processor. For example, the first processor performs step A, the second processor performs step B, or the first processor and the second processor jointly perform steps A and B.

[0129] Furthermore, this embodiment of the invention also provides a readable storage medium, wherein computer-executable instructions are preset in the readable storage medium, and when the processor executes the computer-executable instructions, the above-mentioned reservoir scheduling and management method is implemented.

[0130] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A reservoir operation management method characterized by comprising: The method comprises: constructing a reservoir scheduling entity association graph, the reservoir scheduling entity association graph comprising reservoir entity nodes, runoff input nodes, water demand nodes and scheduling rule nodes, each node establishing an influence conduction path through a directed edge, and the influence conduction path being marked with a path influence weight; performing dynamic evolution update on the reservoir scheduling entity association graph based on hydrological time series data, updating real-time state attributes of each node through a node state propagation algorithm, adjusting the path influence weight of the influence conduction path according to the real-time state attributes, and generating a time-varying scheduling association graph; performing multi-agent collaborative reasoning on the time-varying scheduling association graph, performing deep reasoning on the entity association relationship in the time-varying scheduling association graph through a reservoir scheduling reasoning model, obtaining an influence degree matrix and a scheduling priority sequence between entities; establishing a reservoir joint scheduling scheme according to the influence degree matrix and the scheduling priority sequence, the reservoir joint scheduling scheme comprising a water source allocation sub-scheme, a demand response sub-scheme and a rule coordination sub-scheme; generating a cascade scheduling instruction set based on the reservoir joint scheduling scheme, and sending the cascade scheduling instruction set to a reservoir control center to control reservoir gate regulation, unit output adjustment and water delivery flow distribution.

2. The reservoir operation management method according to claim 1, characterized by, The method comprises: determining the node types of the reservoir scheduling entity association graph, the node types comprising reservoir entity nodes, runoff input nodes, water demand nodes and scheduling rule nodes, wherein the reservoir entity nodes are used to represent physical state information of a single reservoir, the runoff input nodes are used to represent natural inflow information of the reservoir, the water demand nodes are used to represent user demand information dependent on the water supply of the reservoir, and the scheduling rule nodes are used to represent constraint condition information guiding the scheduling of the reservoir; extracting core attribute parameters of each node, for the reservoir entity nodes, the core attribute parameters include the current water level, the effective reservoir capacity, the inflow, the outflow and the flood discharge capacity; for the runoff input nodes, the core attribute parameters include the runoff source type, the runoff process form and the runoff prediction period length; for the water demand nodes, the core attribute parameters include the demand type, the demand time period distribution and the water demand flow size; and for the scheduling rule nodes, the core attribute parameters include the rule applicable condition, the constraint index type and the index control range; establishing a directed edge connection relationship between the nodes, the reservoir entity nodes and the runoff input nodes form an input association through a directed edge, the reservoir entity nodes and the water demand nodes form an output association through a directed edge, and the scheduling rule nodes form a constraint association with the reservoir entity nodes, the runoff input nodes and the water demand nodes through a directed edge; labeling initial path influence weights for each influence conduction path, the initial path influence weights being determined based on the actual influence degree between the nodes, the path influence weight of the runoff input nodes on the reservoir entity nodes being determined according to the correlation analysis of historical runoff and reservoir inflow, the path influence weight of the water demand nodes on the reservoir entity nodes being determined according to the water demand guarantee priority, and the path influence weight of the scheduling rule nodes on other nodes being determined according to the mandatory degree of the rules.

3. The reservoir operation management method according to claim 1, characterized by, The hydrological time series data is used to update the reservoir scheduling entity association graph dynamically, the node state propagation algorithm is used to update the real-time state attributes of each node, the path influence weight of the influence conduction path is adjusted according to the real-time state attributes, and a time-varying scheduling association graph is generated, including: A hydrological time series data sequence is collected, the hydrological time series data sequence includes real-time monitoring data of a reservoir entity node, forecast data of a runoff input node, real-time feedback data of a water demand node and updated data of a scheduling rule node; Propagation parameters of the node state propagation algorithm are initialized, the propagation parameters include a propagation step, a decay coefficient and a convergence threshold, the propagation step is determined according to a decision cycle of reservoir scheduling, and the decay coefficient is used to control the influence intensity attenuation of a remote node; The initial core attribute parameters of each node are input into the node state propagation algorithm, in the propagation process, each node receives a state influence value transmitted by a predecessor node through an influence conduction path, and the state influence value is a product of the core attribute parameters of the predecessor node and the path influence weight of the corresponding influence conduction path; All state influence values received by each node are accumulated, and a state update increment of the node is calculated in combination with the decay coefficient, and the propagation is stopped when the state update increment of K consecutive iterations is less than the convergence threshold; The path influence weight of each influence conduction path is recalculated based on the updated real-time state attributes, for the path from the runoff input node to the reservoir entity node, the path influence weight is adjusted according to the current runoff forecast accuracy, for the path from the water demand node to the reservoir entity node, the path influence weight is adjusted according to the current water demand urgency, and for the path from the scheduling rule node to other nodes, the path influence weight is adjusted according to the current applicability of the rule; The updated real-time state attributes and the adjusted path influence weight are applied to the reservoir scheduling entity association graph, and a time-varying scheduling association graph that dynamically changes with time is generated.

4. The reservoir operation management method according to claim 1, characterized by, Multi-agent collaborative reasoning is performed on the time-varying scheduling association graph, a reservoir scheduling reasoning model is used to deeply reason the entity association relationship in the time-varying scheduling association graph, and an influence degree matrix and a scheduling priority sequence between entities are obtained, including: A reasoning framework of the reservoir scheduling reasoning model is constructed, the reasoning framework includes an entity association extraction layer, a relationship reasoning layer and a priority sorting layer, the entity association extraction layer is used to extract direct association features between entities in the time-varying scheduling association graph, the relationship reasoning layer is used to reason indirect influence relationships between entities, and the priority sorting layer is used to determine a scheduling processing sequence of each node; The node real-time state attribute matrix and the path influence weight adjacency matrix of the time-varying scheduling association graph are input into the entity association extraction layer, the direct association strength between nodes is calculated through the graph attention mechanism of the entity association extraction layer, a direct association feature matrix is generated, and elements in the direct association feature matrix represent the direct influence degree between corresponding two nodes; The direct association feature matrix is input into the relationship reasoning layer, effective indirect influence paths between entities are mined through the path reasoning algorithm of the relationship reasoning layer, and an indirect association feature matrix is generated based on the effective indirect influence paths; The direct association feature matrix and the indirect association feature matrix are fused to generate a comprehensive association feature matrix, the comprehensive association feature matrix is input into an influence degree calculation module of a relationship reasoning layer, the comprehensive association feature matrix is decomposed into an entity influence vector through a matrix decomposition algorithm, and an influence degree matrix between entities is calculated based on the entity influence vector, wherein a row vector in the influence degree matrix represents a comprehensive influence degree of a corresponding node on all other nodes; The influence degree matrix is input into a priority sorting layer, each node is comprehensively scored through a multi-factor decision algorithm of the priority sorting layer, the multi-factor decision algorithm considers the influence range, influence strength and current state urgency of the node, the nodes are sorted from high to low according to the comprehensive scores, and a scheduling priority sequence is generated.

5. The reservoir operation management method according to claim 4, characterized by, The direct association strength between nodes is calculated through a graph attention mechanism of the entity association extraction layer, and a direct association feature matrix is generated, including: The node real-time state attribute matrix is linearly transformed to generate a query feature matrix, a key feature matrix and a value feature matrix, the query feature matrix and the key feature matrix are used to calculate attention scores, and the value feature matrix is used to generate association features; For each node in the time-varying scheduling association graph, dot product operations are performed on a corresponding row vector of the query feature matrix and all column vectors of the key feature matrix to obtain original attention scores of the node on all other nodes; The original attention scores are exponentially processed, and attention weights are obtained through row normalization, wherein the attention weights reflect the importance of the direct association between nodes; The attention weights and the value feature matrix are subjected to weighted summation operation to generate an association feature vector of each node, the association feature vectors of all nodes are arranged in rows to generate the direct association feature matrix, and the number of rows and the number of columns of the direct association feature matrix are equal to the total number of nodes.

6. The reservoir operation management method according to claim 4, characterized by, The direct association feature matrix is input into the relationship reasoning layer, an effective indirect influence path between entities is mined through a path reasoning algorithm of the relationship reasoning layer, and an indirect association feature matrix is generated based on the effective indirect influence path, including: The path length range and the weight threshold value of the path reasoning algorithm are initialized; For each node pair in the time-varying scheduling association graph, all directed paths with lengths within the path length range are extracted, each directed path is composed of continuous influence transmission paths; The cumulative influence weight of each directed path is calculated, and the cumulative influence weight is the product of the path influence weights of all influence transmission paths in the directed path; The directed paths with cumulative influence weights greater than the weight threshold value are marked as effective indirect influence paths, and the starting node, the terminal node and the cumulative influence weight of the effective indirect influence path are recorded; For each node pair, the cumulative influence weights of all effective indirect influence paths are summarized, the summary result is used as the indirect influence strength of the node pair, and an indirect association feature matrix is generated based on the indirect influence strengths of all node pairs, wherein the number of rows and the number of columns of the indirect association feature matrix are equal to the total number of nodes.

7. The reservoir operation management method according to claim 1, characterized by, The reservoir joint scheduling scheme is established according to the influence degree matrix and the scheduling priority sequence, including: Analyzing a reservoir entity node row vector in the influence degree matrix, extracting an influence degree value of each reservoir entity node to a runoff input node, combining a runoff input node sequence in the scheduling priority sequence, determining a water source contribution weight of each runoff input node, and the water source contribution weight is positively correlated with the influence degree value; Based on the water source contribution weight, a water source allocation sub-scheme is established, and the water source allocation sub-scheme includes a runoff input node to a reservoir entity node water inflow allocation ratio and a time period allocation plan; Analyzing a water demand node column vector in the influence degree matrix, extracting an influence degree value of each reservoir entity node to a water demand node, combining a water demand node sequence in the scheduling priority sequence, and constructing a water demand satisfaction degree matrix, an element in the water demand satisfaction degree matrix represents a water supply guarantee capability of a corresponding reservoir entity node to a water demand node; Based on the water demand satisfaction degree matrix, a demand response sub-scheme is established, and the demand response sub-scheme includes a water supply source allocation, a water supply time period arrangement and a water supply flow adjustment plan of each water demand node; Analyzing a related row vector of a scheduling rule node in the influence degree matrix, extracting an influence degree value of the scheduling rule node to other nodes, combining a scheduling rule node sequence in the scheduling priority sequence, and determining a constraint intensity coefficient of each scheduling rule, and the constraint intensity coefficient is positively correlated with the influence degree value; Based on the constraint intensity coefficient, a rule coordination sub-scheme is established, and the rule coordination sub-scheme includes an applicable priority order of the scheduling rule, a coordination strategy of the conflict rule and a dynamic adjustment range of the rule index; Integrating the water source allocation sub-scheme, the demand response sub-scheme and the rule coordination sub-scheme, a reservoir joint scheduling scheme including multi-dimensional scheduling targets is generated.

8. The reservoir operation management method according to claim 7, characterized by, The demand response sub-scheme based on the water demand satisfaction degree matrix includes: Performing row normalization processing on the water demand satisfaction degree matrix to obtain a water supply allocation ratio of each reservoir entity node to a water demand node, and the water supply allocation ratio represents a water supply contribution share of the reservoir entity node to the water demand node; According to a water demand node sequence in the scheduling priority sequence, a water supply source is allocated to each water demand node in turn; Analyzing a water demand time period distribution characteristic of the water demand node, combining an inflow process of the reservoir entity node and an outflow capacity constraint, establishing a water supply time period arrangement for each water demand node, and the water supply time period arrangement matches a water demand peak time period of the water demand node with a water supply capacity peak time period of the reservoir entity node; Based on a water demand flow size of the water demand node and the water supply allocation ratio, an initial water supply flow of each reservoir entity node to the water demand node is calculated, the initial water supply flow is adjusted in combination with a current water level and a reservoir capacity state of the reservoir entity node, and a water supply flow adjustment plan is generated; Integrating the water supply source allocation, the water supply time period arrangement and the water supply flow adjustment plan, a complete demand response sub-scheme is generated, and the demand response sub-scheme also includes a flow monitoring point setting and a deviation adjustment mechanism in a water supply process, when a deviation between an actual water supply flow and a planned water supply flow exceeds a preset range, the deviation adjustment mechanism is started to dynamically correct the water supply flow.

9. The reservoir operation management method according to claim 1, characterized by, The reservoir joint dispatching scheme is based on generating a cascade scheduling instruction set, sending the cascade scheduling instruction set to the reservoir control center to control the reservoir gate regulation, unit output adjustment and water distribution, comprising: Analyzing the water source allocation sub-scheme in the reservoir joint dispatching scheme, extracting the inflow forecast value and time period allocation plan of each reservoir entity node, combining the current water level and reservoir capacity curve of the reservoir entity node, calculating the target water storage level of each period; Based on the target water storage level, the reservoir gate regulation instruction is generated, which contains the gate type, regulation period, target opening degree and opening degree change rate. For the flood discharge gate, the number of gate opening groups and the opening degree are determined according to the inflow forecast value and the flood control limit water level; Analyzing the demand response sub-scheme in the reservoir joint dispatching scheme, extracting the water supply flow regulation plan of each reservoir entity node, combining the power generation water head characteristic curve of the reservoir entity node, calculating the target power generation output of each period; Based on the target power generation output, the unit output adjustment instruction is generated, which contains the unit number, the number of starting stations, the target output and the output adjustment rate. According to the priority relationship between the water supply flow regulation plan and the target power generation output, the flow distribution of power generation water and other water is coordinated; Analyzing the rule coordination sub-scheme in the reservoir joint dispatching scheme, extracting the dynamic adjustment range of the rule index of each reservoir entity node, combining the hydraulic connection relationship of the cross-reservoir water conveyance system, calculating the target water conveyance flow of each period; Based on the target water conveyance flow, the water conveyance flow distribution instruction is generated, which contains the water conveyance channel number, the target flow, the regulation period and the operation plan of the regulating gate along the line. According to the joint operation requirements of cascade reservoirs, the water conveyance flow connection of upstream reservoirs and downstream reservoirs is coordinated; Integrate the reservoir gate regulation instruction, the unit output adjustment instruction and the water conveyance flow distribution instruction to generate a cascade scheduling instruction set containing time stamp alignment. Each cascade scheduling instruction in the cascade scheduling instruction set contains execution subject identification, instruction effective time, target parameter and operation constraint condition; The cascade scheduling instruction set is sent to the reservoir control center, so that the reservoir control center controls the gate actuator, generator unit and water conveyance control equipment of each reservoir according to the cascade scheduling instruction set.

10. A reservoir operation management system characterized by comprising: It includes a processor and a memory, the memory and the processor are connected, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to realize the reservoir dispatching management method in any one of claims 1-9.

Images