Power system flexible resource scheduling control method and system based on graph database

Abstract

The application discloses a power system flexible resource scheduling control method and system based on a graph database, relates to the technical field of resource scheduling control, and comprises the following steps: constructing a power system topology graph model based on a graph database, and dividing an initial section boundary region through a shortest path algorithm; constructing a dynamic boundary model, and identifying an abnormal section region in combination with a preset power flow equation; acquiring a candidate capacity boundary set; under a plurality of preset disturbance scales, filtering and disturbing the candidate capacity boundary set to obtain a plurality of capacity response values, and determining an optimal credible capacity boundary; updating the optimal credible capacity boundary to the graph database and generating a core flexible resource scheduling instruction to perform resource scheduling. The application solves the problem that, in traditional scheduling, a closed-loop mechanism from capacity boundary analysis to graph database updating and scheduling instruction generation is not established, so that capacity boundary information cannot effectively guide scheduling decisions.

Description

Technical Field

[0001] This invention relates to the field of resource scheduling and control technology, and more specifically, to a flexible resource scheduling and control method and system for power systems based on graph databases. Background Technology

[0002] Flexible resource dispatching in the power system is a crucial link in ensuring the safe and stable operation of the power grid, undertaking the core tasks of coordinating generation and load, maintaining power balance in the grid, and improving system response speed and stability. With the large-scale grid connection of renewable energy sources such as wind and solar power, the power grid operating environment has become increasingly complex and volatile, significantly increasing system volatility and uncertainty. Furthermore, the diversified structure of modern power loads, such as electric vehicle charging, large-scale energy storage facilities, and distributed power sources, further exacerbates the operational complexity and dispatching difficulties of the power system.

[0003] Traditional scheduling methods mainly rely on static models and empirical rules, using linear programming or heuristic algorithms for capacity allocation and scheduling decisions. However, these methods usually ignore the dynamic changes in the power system topology and the time-varying characteristics of abnormal cross-sectional areas, making it difficult to fully reflect the actual operating conditions of the system, resulting in insufficient robustness and accuracy of the scheduling scheme.

[0004] For example, the invention patent announcement CN115514014B discloses a novel power system flexibility resource supply and demand game optimization scheduling method with a high proportion of wind power. This method includes establishing a two-layer game architecture for flexibility resource supply and demand based on a master-slave game; establishing a quantitative model of wind power operators' flexibility resource demand based on wind power volatility indicators; establishing an incentive price optimization decision model for upper-level wind power operators towards lower-level energy storage operators, thermal power operators, and demand response aggregators; establishing a flexibility resource supply decision model for lower-level energy storage operators, thermal power operators, and demand response aggregators; combining the established incentive price optimization decision model and the established flexibility resource supply decision model to jointly constitute a novel power system flexibility resource supply and demand game optimization model with a high proportion of wind power, and then solving it. This invention can effectively improve the enthusiasm of multiple parties involved in the regulation of power system flexibility resources (source-load-storage) and promote the grid connection and consumption of high-proportion wind power.

[0005] The above-disclosed technical solutions have at least the following technical problems:

[0006] Traditional flexible resource scheduling and control methods for power systems lack a closed-loop mechanism from capacity boundary analysis to graph database updates and scheduling command generation. This results in capacity boundary information failing to effectively guide scheduling decisions, reducing scheduling accuracy and the system's adaptability. To address these issues, this invention proposes a solution. Summary of the Invention

[0007] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a flexible resource scheduling and control method and system for power systems based on graph databases. By constructing a dynamic topology model through a graph database, combining an improved shortest path algorithm to divide the initial cross-section, and determining robust capacity boundaries through multi-objective optimization and disturbance testing, adaptive scheduling instructions are finally generated. This addresses the problem in traditional scheduling where there is no closed-loop mechanism from capacity boundary analysis to graph database updates and scheduling instruction generation, resulting in capacity boundary information being unable to effectively guide scheduling decisions.

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] A flexible resource scheduling and control method and system for power systems based on graph databases includes the following steps:

[0010] A power system topology model is constructed based on a graph database, and the initial cross-sectional boundary regions are delineated using a shortest path algorithm. A dynamic boundary model is then constructed based on these initial boundary regions, and abnormal cross-sectional regions are identified using pre-defined power flow equations. A multi-objective optimization function is constructed for these abnormal cross-sectional regions and solved using a non-dominated sorting genetic algorithm to obtain a set of candidate capacity boundaries. Under several pre-defined perturbation scales, the candidate capacity boundary sets are filtered to obtain several capacity response values. A reliability evaluation function is then constructed based on the matching degree between the capacity response values ​​and the pre-defined normal system capacity threshold to determine the optimal reliable capacity boundary. Finally, the optimal reliable capacity boundary is updated to the graph database, and flexible resource scheduling instructions are generated for resource scheduling.

[0011] In a preferred embodiment, the step of constructing a power system topology graph model based on a graph database and dividing the initial cross-sectional boundary region using a shortest path algorithm specifically involves: acquiring equipment attribute data for all nodes in the power system, including generator capacity, load power, line impedance, and node voltage level; mapping nodes to vertices in the graph database, mapping connections to edges, and hierarchically labeling vertices based on node voltage level to construct a power system topology graph model; using Dijkstra's algorithm to traverse the shortest paths in the power system topology graph model to obtain several first shortest paths; obtaining path feature sequences of the first shortest paths based on the equipment attribute data to obtain several path feature sequences; using the DTW algorithm to filter the several path feature sequences to obtain several second shortest paths; and performing cluster analysis on the several second shortest paths to divide the initial cross-sectional boundary region.

[0012] In a preferred embodiment, the step of performing cluster analysis on several second shortest paths to delineate the initial cross-sectional boundary region specifically involves: constructing a first path feature vector from the node and device attribute data of each second shortest path; clustering the several second shortest paths using a clustering algorithm based on the first path feature vector to obtain several path clusters; constructing a node frequency vector and an edge frequency matrix within each path cluster; determining whether each edge endpoint crosses a cluster based on the node frequency vector and the edge frequency matrix, and counting the number of times the endpoint appears in the path cluster; if the frequency exceeds a preset threshold, it is marked as a cross-sectional boundary edge; and constructing the initial cross-sectional boundary region from the endpoints of all cross-sectional boundary edges.

[0013] In a preferred embodiment, the step of constructing a dynamic boundary model based on the initial cross-sectional boundary region and identifying abnormal cross-sectional regions in conjunction with a preset power flow equation specifically involves: constructing a local response map model of the initial cross-sectional boundary region and constructing a node power disturbance model based on the preset power flow equation; calculating the disturbance response value of each node in the local response map model based on the node power disturbance model to obtain the disturbance propagation matrix of each node; predicting the power of each node using a preset neural network model to obtain the predicted power value; calculating the residual between the real-time power data and the predicted power value of each node to obtain the power residual value of each node; constructing a first power disturbance vector based on the power residual value of each node and inputting it into the disturbance propagation matrix to obtain the response value vector of the local response map model; and performing data analysis on the response value vector of the local response map model to identify abnormal cross-sectional regions.

[0014] In a preferred embodiment, the step of constructing a multi-objective optimization function for the abnormal cross-sectional region and solving it using a non-dominated sorting genetic algorithm to obtain a candidate capacity boundary set specifically involves: calculating the unit capacity compensation efficiency of the control nodes for the abnormal cross-sectional region based on the abnormal cross-sectional region and combining the power residual values ​​of the nodes, and constructing a regional capacity response matrix; filtering the nodes based on the regional capacity response matrix to obtain the filtered nodes, and using the preset maximum allowable adjustment capacity of the filtered nodes as the upper limit of the capacity boundary to obtain a capacity boundary variable set and optimizing the capacity boundary variable set.

[0015] In a preferred embodiment, the optimization of the capacity boundary variable set specifically involves: constructing a multi-objective optimization function with the optimization objectives of power balance restoration in the abnormal cross-section area, maximizing branch capacity utilization, and minimizing adjustment costs; and optimizing the capacity boundary variable set based on the multi-objective optimization function using a non-dominated sorting strategy to obtain a candidate capacity boundary set, wherein the non-dominated sorting strategy includes variable initialization.

[0016] In a preferred embodiment, the step of filtering and perturbing the candidate capacity boundary set under a preset number of disturbance scales to obtain a number of capacity response values ​​specifically involves: acquiring historical fault data and defining a disturbance type set, which includes load surges, generator disconnection, and line disconnection; setting the disturbance scale level for each disturbance type and constructing a disturbance scenario library; for each disturbance scenario in the disturbance scenario library, randomly selecting a disturbance source node and injecting it into each candidate capacity boundary in the candidate capacity boundary set to obtain a number of disturbance capacity boundaries; and then using the number of disturbance capacity boundaries based on a preset power flow response model to obtain a number of capacity response values.

[0017] In a preferred embodiment, the step of determining the optimal reliable capacity boundary by constructing a reliability evaluation function based on the matching degree between the capacity response value and the preset system normal capacity threshold specifically involves: constructing a capacity response vector for each capacity response value, and performing cluster analysis on the capacity response vector using a clustering algorithm to construct a multi-scale response topology cluster map; constructing a reliability density function based on the multi-scale response topology cluster map using a distribution fitting algorithm; calculating the expected value of each perturbation capacity boundary based on the reliability density function, and selecting the perturbation capacity boundary with the largest expected value as the optimal reliable capacity boundary.

[0018] In a preferred embodiment, updating the optimal trusted capacity boundary to the graph database and generating flexible resource scheduling instructions for resource scheduling specifically involves: parsing the optimal trusted capacity boundary and mapping the parsed parameters to the vertex attributes of the graph database to obtain an updated graph database; constructing a resource scheduling instruction generation model based on the updated graph database and extracting the dependencies and priorities between nodes in the model using a graph neural network; generating a scheduling instruction sequence based on the dependencies and priorities, issuing the scheduling instructions to the power system control terminal, monitoring the instruction execution effect in real time, and dynamically adjusting the capacity boundary parameters in the graph database based on feedback data.

[0019] The technical effects and advantages of the flexible resource scheduling and control method and system for power systems based on graph databases in this invention are as follows:

[0020] 1. This invention constructs a power system topology model in a graph database, combining shortest path partitioning and dynamic boundary modeling to effectively characterize the dynamic changes in the system's operational structure. Based on this, power flow disturbance analysis and residual calculation are introduced to identify abnormal cross-sectional areas, achieving precise anomaly localization. Subsequently, a multi-objective optimization algorithm is used to construct an optimal scheduling model covering power balance, capacity utilization, and cost control, obtaining a candidate capacity boundary set. Filtered responses are then applied under multiple disturbance scenarios to construct a reliability evaluation function, ultimately determining the optimal reliable capacity boundary. This capacity boundary not only possesses high dynamic adaptability and response stability but is also updated in real-time to the graph database. A graph neural network model extracts node dependencies, generating core scheduling instructions including generation, load, and line regulation, completing a closed-loop process from analysis to control. This method overcomes the deficiency in traditional scheduling methods where capacity boundary analysis results cannot be fed back to actual scheduling control, significantly improving the scientific nature, real-time performance, and robustness of scheduling decisions, meeting the higher requirements of new power systems for flexibility, efficiency, and intelligence. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the flexible resource scheduling and control method for power systems based on graph databases according to the present invention.

[0022] Figure 2 This is a schematic diagram of the structure of the flexible resource scheduling and control system for power systems based on graph databases according to the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0024] Example 1, Figure 1 The present invention provides a flexible resource scheduling and control method for power systems based on graph databases, comprising the following steps:

[0025] S1. Construct a power system topology graph model based on a graph database, and divide the initial cross-sectional boundary region using the shortest path algorithm.

[0026] In this example, a power system topology graph model is constructed based on a graph database, and the initial cross-sectional boundary region is divided using a shortest path algorithm, specifically:

[0027] Obtain equipment attribute data for all nodes in the power system, including generator capacity, load power, line impedance, and node voltage level;

[0028] The nodes are mapped to vertices in the graph database, the connections are mapped to edges, and the vertices are hierarchically labeled based on the node voltage level to construct a power system topology graph model.

[0029] Dijkstra's algorithm is used to traverse the shortest paths in the power system topology model to obtain several first shortest paths;

[0030] Based on device attribute data, the path feature sequence of the first shortest path is obtained, resulting in several path feature sequences;

[0031] The DTW algorithm is used to filter several path feature sequences to obtain several second shortest paths;

[0032] Cluster analysis is performed on several second shortest paths to delineate the initial cross-sectional boundary region.

[0033] It should be noted that the following grid node equipment attributes are obtained by extracting system operation data from the power grid management system (such as EMS):

[0034] Generator capacity (MW): For example, node G1 is 300MW, and node G2 is 500MW;

[0035] Load power (MW): For example, load node L1 is 120MW, and L2 is 80MW;

[0036] Node voltage levels (kV): For example, N1 is 220kV, N2 is 110kV, and N3 is 35kV.

[0037] Based on the above data, the following modeling process is performed in the Neo4j graph database:

[0038] Nodes are mapped to vertices: all nodes such as generators, loads, and substations are mapped to vertices of the graph;

[0039] Connection relationships are mapped to edges: For example, if a line connects N1 and N2, then a directed edge from N1 to N2 is created in the graph;

[0040] Voltage level hierarchical labeling: Level labels are added to vertex attribute fields, with 220kV, 110kV, etc. as levels, to facilitate hierarchical processing. At this point, a power topology graph model containing equipment characteristics and connection relationships is formed in the graph database.

[0041] Furthermore, the Dijkstra shortest path algorithm built into the graph database is used to traverse the paths between any generator node and load node, obtaining several first shortest paths. Based on equipment attribute data, the path feature sequences of these first shortest paths are obtained, resulting in several path feature sequences. The DTW (Dynamic Time Warping) algorithm is then used to compare the similarity between all path feature sequences, filtering out distorted paths and retaining a batch of paths with high similarity as the second shortest path set.

[0042] In this example, cluster analysis is performed on several second shortest paths to delineate the initial cross-sectional boundary region, specifically as follows:

[0043] Construct the first path feature vector from the node and device attribute data in each second shortest path;

[0044] Based on the feature vector of the first path, a clustering algorithm is used to cluster several second shortest paths to obtain several path clusters;

[0045] Within each path cluster, construct the node frequency vector and the edge frequency matrix;

[0046] The node frequency vector and edge frequency matrix are used to determine whether each edge endpoint crosses a cluster and to count the number of times the endpoint appears in the path cluster. If the frequency exceeds a preset threshold, it is marked as a cross-sectional boundary edge.

[0047] The endpoints of all cross-sectional boundary edges constitute the initial cross-sectional boundary region.

[0048] It should be noted that each second shortest path is constructed as a path feature vector, with feature dimensions including: voltage level of each node; node type (generator / load / relay); impedance and flow direction of adjacent edges. K-means or DBSCAN clustering algorithms are used to cluster the path feature vectors, dividing them into several path clusters. Within each cluster, the following statistics are performed:

[0049] Node frequency vector: the number of times each node appears in the cluster;

[0050] Edge frequency matrix: the number of times each edge appears in the path.

[0051] The algorithm identifies whether the two endpoints of each edge belong to different clusters. If the two endpoints of an edge cross different clusters and the edge appears more frequently in the path cluster than a set threshold (e.g., 30%), then the edge is marked as a "section boundary edge". Finally, the set of endpoint nodes of all section boundary edges constitutes the initial section boundary region, which serves as the basis for subsequent dynamic modeling and optimization scheduling.

[0052] S2, construct a dynamic boundary model based on the initial cross-sectional boundary region, and identify abnormal cross-sectional regions by combining the preset power flow equations;

[0053] In this example, a dynamic boundary model is constructed based on the initial cross-sectional boundary region, and abnormal cross-sectional regions are identified in conjunction with a preset power flow equation, specifically:

[0054] The initial cross-sectional boundary region is constructed as a local response map model, and a nodal power perturbation model is constructed based on the preset power flow equations.

[0055] The perturbation response value of each node in the local response graph model is calculated based on the node power perturbation model, and the perturbation propagation matrix of each node is obtained.

[0056] The power of each node is predicted using a pre-defined neural network model to obtain the predicted power value;

[0057] The residual between the real-time power data and the predicted power value of each node is calculated to obtain the power residual value of each node;

[0058] The first power perturbation vector is constructed based on the power residual value of each node and input into the perturbation propagation matrix to obtain the response value vector of the local response map model;

[0059] Data analysis is performed on the response value vectors of the local response map model to identify abnormal cross-sectional areas.

[0060] It should be noted that constructing a local response graph model for the initial cross-sectional boundary region involves first extracting all nodes and their connections from the graph database to form a local subgraph, while retaining node attributes such as power information, voltage level, and equipment type. Then, a node power disturbance model based on power flow equations is introduced into this subgraph. By injecting a certain disturbance (such as power change) into each node, its impact on adjacent nodes is calculated, forming a disturbance propagation matrix. Subsequently, a neural network (such as LSTM) is used to predict the power value of each node under normal conditions, and residual analysis is performed between the predicted values ​​and the actual real-time power data to obtain the power residual vector for each node. This residual vector is input into the disturbance propagation matrix to deduce the response value vector of each node in the entire local response graph. Finally, statistical analysis methods (such as PCA dimensionality reduction and cluster analysis) are used to determine which nodes' response values ​​deviate significantly under disturbance. If multiple high-response-value nodes are concentrated in a certain area, that area can be identified as an abnormal cross-sectional region, indicating a potential risk of power flow imbalance or abnormal power propagation.

[0061] S3. Construct a multi-objective optimization function for the abnormal cross-sectional area and solve it using a non-dominated sorting genetic algorithm to obtain a candidate capacity boundary set;

[0062] In this example, a multi-objective optimization function is constructed for the abnormal cross-sectional region and solved using a non-dominated sorting genetic algorithm to obtain a candidate capacity boundary set, specifically:

[0063] Based on the abnormal cross-sectional area, the pre-set control node's unit capacity compensation efficiency for the abnormal cross-sectional area is calculated by combining the power residual value of the node, and the regional capacity response matrix is ​​constructed.

[0064] The nodes are filtered based on the regional capacity response matrix to obtain the filtered nodes, and the preset maximum allowable adjustment capacity of the filtered nodes is used as the upper limit of the capacity boundary to obtain the capacity boundary variable set.

[0065] A multi-objective optimization function is constructed with the optimization objectives of restoring power balance in abnormal cross-section areas, maximizing branch capacity utilization, and minimizing regulation costs.

[0066] Set the population size and number of iterations for the genetic algorithm, and initialize decision variables including generator output adjustment, load transfer, and energy storage scheduling;

[0067] A candidate capacity boundary set is obtained by using a multi-objective optimization function.

[0068] It should be noted that control nodes refer to key nodes with adjustment capabilities that can be used to restore system power balance and alleviate load pressure in abnormal areas. These typically include generators, controllable loads, and energy storage devices, which participate in scheduling as potential intervention measures during the optimization process. Unit capacity compensation efficiency is an indicator that measures the effect of a control node's unit capacity adjustment on power restoration in abnormal cross-sectional areas. It is usually calculated by analyzing the degree of regional response caused by the unit output change of the node in the disturbance propagation matrix, reflecting its marginal benefit in alleviating local anomalies. Based on these control nodes and their unit capacity compensation efficiency, the constructed multi-objective optimization function comprehensively considers multiple optimization objectives, including power balance restoration in abnormal areas (i.e., minimizing the total residual), maximizing branch capacity utilization (minimizing idle or overloaded lines), and minimizing adjustment costs (economic constraints of control nodes). By establishing trade-offs between these objectives, a Pareto optimal solution set is generated using a non-dominated sorting strategy in multi-objective solution methods such as genetic algorithms, ultimately providing the system with multiple scheduling strategy candidate schemes.

[0069] S4. Under several preset perturbation scales, filter and perturb the candidate capacity boundary set to obtain several capacity response values, and construct a credibility evaluation function based on the matching degree between the capacity response values ​​and the preset system normal capacity threshold to determine the optimal credibility capacity boundary.

[0070] In this example, several capacity response values ​​are obtained by filtering and perturbing the candidate capacity boundary set under several preset perturbation scales, specifically:

[0071] Acquire historical fault data and define a set of disturbance types, which includes load surges, generator disconnection from the grid, and line breaks;

[0072] Define the perturbation scale level for each perturbation type and build a perturbation scenario library;

[0073] For each disturbance scenario in the disturbance scenario library, a disturbance source node is randomly selected and injected into each candidate capacity boundary in the candidate capacity boundary set, resulting in several disturbance capacity boundaries;

[0074] Several capacity response values ​​are obtained by basing several disturbance capacity boundaries on a pre-defined power flow response model.

[0075] It should be noted that, in order to comprehensively evaluate the reliability and robustness of candidate capacity boundaries in actual operation, historical fault data was first collected and organized, and various disturbance types, including load surges, generator disconnection, and line breaks, were defined to construct a disturbance scenario library covering different disturbance scales and scenarios. For each disturbance type, multiple disturbance scale levels were set to reflect different disturbance intensities, thereby simulating the power grid response under various abnormal conditions. During the simulation, disturbance source nodes were randomly selected and the disturbances were injected into each capacity boundary scheme in the candidate capacity boundary set, generating a series of disturbance capacity boundaries. Subsequently, based on the preset power flow response model, the power flow changes in the power grid for each disturbance capacity boundary were calculated to obtain the corresponding capacity response values. These response values ​​reflect the operating performance of the power grid under different disturbances, thus providing key data support for credibility evaluation and selection of the optimal capacity boundary.

[0076] In this example, a reliability evaluation function is constructed based on the matching degree between the capacity response value and the preset system normal capacity threshold to determine the optimal reliable capacity boundary, specifically:

[0077] For each capacity response value, a capacity response vector is constructed, and a clustering algorithm is used to perform cluster analysis on the capacity response vector to construct a multi-scale response topology clustering map;

[0078] Based on the multi-scale response topological clustering map, a confidence density function is constructed using a distribution fitting algorithm;

[0079] The expected value of each perturbation capacity boundary is calculated based on the confidence density function, and the perturbation capacity boundary with the largest expected value is selected as the optimal confidence capacity boundary.

[0080] It should be noted that, for multiple capacity response values ​​obtained under different disturbance scenarios, each response value is first constructed into a capacity response vector, and these vectors are analyzed using a clustering algorithm to form a multi-scale response topology cluster map, thereby revealing the distribution characteristics and intrinsic correlations of the capacity response under different disturbance conditions. Based on this cluster map, a distribution fitting algorithm is further used to establish a reliability density function to quantify the probability that each capacity response scheme meets the normal capacity threshold of the system in actual operation. By calculating the expected value of the reliability density function corresponding to each disturbance capacity boundary, its overall performance and stability are evaluated. Finally, the disturbance capacity boundary with the highest expected value is selected as the optimal reliable capacity boundary, thereby ensuring that the flexible resource scheduling scheme has high reliability and adaptability under various disturbance conditions.

[0081] S5 updates the optimal trusted capacity boundary to the graph database and generates flexible resource scheduling instructions for resource scheduling.

[0082] In this example, the optimal trusted capacity boundary is updated to the graph database and flexible resource scheduling instructions are generated for resource scheduling, specifically as follows:

[0083] Analyze generator output adjustment values, load allocation strategies, and line switching schemes within the optimal reliable capacity boundary;

[0084] The parsed parameters are mapped to the vertex attributes of the graph database, and the node capacity, edge transmission limit and topology connection state are updated.

[0085] Based on the updated graph database, a resource scheduling instruction generation model is constructed. The model extracts the dependencies and priorities between nodes through a graph neural network.

[0086] The scheduling instruction sequence is generated based on dependencies and priorities, including generator output instructions, load switching instructions, and standby line switching instructions;

[0087] Dispatch instructions are sent to the power system control terminal, the execution effect of instructions is monitored in real time, and the capacity boundary parameters in the graph database are dynamically adjusted based on feedback data.

[0088] It should be noted that the key scheduling parameters in the optimal reliable capacity boundary—including generator output adjustment values, load allocation strategies, and standby line switching schemes—are analyzed. These parameters are then mapped and updated to the attributes of corresponding nodes and edges in the graph database, such as node capacity, line transmission limits, and topology connection status, thereby enabling dynamic adjustment of the power system model. Based on the updated graph database, a resource scheduling instruction generation model using a graph neural network (GNN) is constructed. The GNN intelligently generates scheduling sequences containing generator output instructions, load switching instructions, and standby line switching instructions by capturing complex dependencies and priorities between nodes. Finally, these scheduling instructions are issued to the power system control terminal, enabling precise scheduling and real-time management of flexible resources. Simultaneously, the system continuously monitors the execution effect of the instructions and dynamically adjusts the capacity boundary parameters in the graph database based on feedback data, ensuring the effectiveness of the scheduling strategy and the stability of system operation.

[0089] Example 2, Figure 2 This invention presents a flexible resource scheduling and control system for power systems based on graph databases, including a boundary partitioning module, an anomaly screening module, an optimization solution module, a disturbance assessment module, and a scheduling and control module.

[0090] The boundary delineation module is used to construct a power system topology graph model based on a graph database and to delineate the initial cross-sectional boundary regions using a shortest path algorithm.

[0091] The anomaly screening module is used to construct a dynamic boundary model based on the initial cross-sectional boundary region and identify abnormal cross-sectional regions in combination with the preset power flow equations.

[0092] The optimization solution module is used to construct a multi-objective optimization function for the abnormal cross-sectional region and solve it using a non-dominated sorting genetic algorithm to obtain a candidate capacity boundary set.

[0093] The disturbance assessment module is used to filter and disturb the candidate capacity boundary set under several preset disturbance scales to obtain several capacity response values, and to construct a confidence evaluation function based on the matching degree between the capacity response values ​​and the preset system normal capacity threshold to determine the optimal confidence capacity boundary.

[0094] The scheduling and control module is used to update the optimal trusted capacity boundary to the graph database and generate flexible resource scheduling instructions for resource scheduling.

[0095] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0096] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0097] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0098] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0099] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0100] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for power system flexible resource dispatching control based on a graph database, characterized in that, Includes the following steps: A power system topology graph model is constructed based on a graph database, and the initial cross-sectional boundary region is divided using the shortest path algorithm. The initial cross-sectional boundary region is constructed as a local response map model, and a nodal power perturbation model is constructed based on the preset power flow equations. The perturbation response value of each node in the local response graph model is calculated based on the node power perturbation model, and the perturbation propagation matrix of each node is obtained. The power of each node is predicted using a pre-defined neural network model to obtain the predicted power value; The residual between the real-time power data and the predicted power value of each node is calculated to obtain the power residual value of each node; The first power perturbation vector is constructed based on the power residual value of each node and input into the perturbation propagation matrix to obtain the response value vector of the local response map model; Data analysis is performed on the response value vectors of the local response map model to identify abnormal cross-sectional areas; A multi-objective optimization function is constructed for the abnormal cross-sectional region and solved by a non-dominated sorting genetic algorithm to obtain a candidate capacity boundary set. Under several preset perturbation scales, the candidate capacity boundary set is filtered and perturbed to obtain several capacity response values. Based on the matching degree between the capacity response values ​​and the preset normal system capacity threshold, a credibility evaluation function is constructed to determine the optimal credibility capacity boundary. The optimal trusted capacity boundary is updated to the graph database and flexible resource scheduling instructions are generated for resource scheduling. 2.The power system flexible resource scheduling control method based on a graph database according to claim 1, characterized in that, The process of constructing a power system topology model based on a graph database and dividing the initial cross-sectional boundary region using a shortest path algorithm is as follows: Obtain equipment attribute data for all nodes in the power system, including generator capacity, load power, line impedance, and node voltage level; The nodes are mapped to vertices in the graph database, the connections are mapped to edges, and the vertices are hierarchically labeled based on the node voltage level to construct a power system topology graph model. Dijkstra's algorithm is used to traverse the shortest paths in the power system topology model to obtain several first shortest paths; Based on device attribute data, the path feature sequence of the first shortest path is obtained, resulting in several path feature sequences; The DTW algorithm is used to filter several path feature sequences to obtain several second shortest paths; Cluster analysis is performed on several second shortest paths to delineate the initial cross-sectional boundary region. 3.The graph database-based power system flexible resource scheduling control method of claim 2, wherein, The clustering analysis of several second shortest paths to delineate the initial cross-sectional boundary region specifically involves: Construct the first path feature vector from the node and device attribute data in each second shortest path; Based on the feature vector of the first path, a clustering algorithm is used to cluster several second shortest paths to obtain several path clusters; Within each path cluster, construct the node frequency vector and the edge frequency matrix; Based on the node frequency vector and the edge frequency matrix, it is determined whether each edge endpoint crosses a cluster, and the number of times the endpoint appears in the path cluster is counted. If the frequency exceeds the preset threshold, it is marked as a cross-sectional boundary edge. The endpoints of all cross-sectional boundary edges constitute the initial cross-sectional boundary region. 4.The graph database-based power system flexible resource scheduling control method of claim 3, wherein, The process of constructing a multi-objective optimization function for the abnormal cross-sectional region and solving it using a non-dominated sorting genetic algorithm to obtain a candidate capacity boundary set is as follows: Based on the abnormal cross-sectional area, the pre-set control node's unit capacity compensation efficiency for the abnormal cross-sectional area is calculated by combining the power residual value of the node, and the regional capacity response matrix is ​​constructed. The nodes are filtered based on the regional capacity response matrix to obtain the filtered nodes. The preset maximum allowable adjustment capacity of the filtered nodes is used as the upper limit of the capacity boundary to obtain the capacity boundary variable set and optimize the capacity boundary variable set.

5. The graph database-based power system flexible resource scheduling control method according to claim 4, characterized in that, The optimization of the capacity boundary variable set specifically involves: A multi-objective optimization function is constructed with the optimization objectives of restoring power balance in abnormal cross-section areas, maximizing branch capacity utilization, and minimizing regulation costs. Based on a multi-objective optimization function, a non-dominated sorting strategy is used to optimize the capacity boundary variable set to obtain a candidate capacity boundary set. The non-dominated sorting strategy includes variable initialization.

6. The graph database-based power system flexible resource scheduling control method according to claim 5, characterized in that, The process of filtering and perturbing the candidate capacity boundary set at several preset perturbation scales to obtain several capacity response values ​​is as follows: Acquire historical fault data and define a set of disturbance types, which includes load surges, generator disconnection from the grid, and line breaks; Define the perturbation scale level for each perturbation type and build a perturbation scenario library; For each disturbance scenario in the disturbance scenario library, a disturbance source node is randomly selected and injected into each candidate capacity boundary in the candidate capacity boundary set, resulting in several disturbance capacity boundaries; Several capacity response values ​​are obtained by basing several disturbance capacity boundaries on a pre-defined power flow response model.

7. The flexible resource scheduling and control method for power systems based on graph databases according to claim 6, characterized in that, The optimal reliable capacity boundary is determined by constructing a reliability evaluation function based on the matching degree between the capacity response value and the preset system normal capacity threshold. Specifically: For each capacity response value, a capacity response vector is constructed, and a clustering algorithm is used to perform cluster analysis on the capacity response vector to construct a multi-scale response topology clustering map; Based on the multi-scale response topological clustering map, a confidence density function is constructed using a distribution fitting algorithm; The expected value of each perturbation capacity boundary is calculated based on the confidence density function, and the perturbation capacity boundary with the largest expected value is selected as the optimal confidence capacity boundary.

8. The flexible resource scheduling and control method for power systems based on graph databases according to claim 7, characterized in that, The step of updating the optimal trusted capacity boundary to the graph database and generating flexible resource scheduling instructions for resource scheduling specifically involves: The optimal reliable capacity boundary is analyzed, and the analyzed parameters are mapped to the vertex attributes of the graph database to obtain the updated graph database. Based on the updated graph database, a resource scheduling instruction generation model is constructed, and the dependency relationships and priorities between nodes in the model are extracted through a graph neural network. The system generates a sequence of scheduling instructions based on dependencies and priorities, sends the instructions to the power system control terminal, monitors the execution effect of the instructions in real time, and dynamically adjusts the capacity boundary parameters in the graph database based on feedback data.

9. A power system flexible resource scheduling and control system based on a graph database, applied to the power system flexible resource scheduling and control method based on a graph database as described in any one of claims 1-8, characterized in that, It includes a boundary partitioning module, an anomaly filtering module, an optimization solution module, a disturbance assessment module, and a scheduling control module. The boundary division module is used to construct a power system topology graph model based on a graph database, divide the initial cross-sectional boundary region using a shortest path algorithm, and output the initial cross-sectional boundary region to the anomaly filtering module. An anomaly screening module is used to receive the initial cross-sectional boundary region and construct it into a local response map model including the node power perturbation model. Based on the power residual value of each node and the perturbation propagation matrix corresponding to the node power perturbation model, the response value vector of the local response map model is calculated. The abnormal cross-sectional region is identified by data analysis of the response value vector, and the abnormal cross-sectional region is output to the optimization solution module. The optimization solution module is used to receive the abnormal cross-sectional area, construct a multi-objective optimization function for the abnormal cross-sectional area and solve it using a non-dominated sorting genetic algorithm to obtain a candidate capacity boundary set, and output the candidate capacity boundary set to the disturbance evaluation module. The disturbance assessment module is used to receive the candidate capacity boundary set, filter and disturb the candidate capacity boundary set under several preset disturbance scales to obtain several capacity response values, construct a credibility evaluation function based on the matching degree between the capacity response values ​​and the preset system normal capacity threshold to determine the optimal credibility capacity boundary, and output the optimal credibility capacity boundary to the scheduling control module. The scheduling control module is used to receive the optimal trusted capacity boundary, update the optimal trusted capacity boundary to the graph database, and generate flexible resource scheduling instructions for resource scheduling.

Images