A data splitting processing method of an electromagnetic simulation model of a power system

Abstract

The application relates to the technical field of power simulation, and discloses a data splitting processing method for an electromagnetic simulation model of a power system. The method comprises the following steps: constructing a node admittance matrix and forming an electrical distance matrix through an eigenvalue decomposition algorithm; performing a K-means clustering algorithm based on the electrical distance matrix to divide nodes into multiple electrically coupled subsystems; identifying boundary connection relationships between the subsystems to determine boundary nodes and boundary branches; performing Thevenin equivalent processing on the boundary nodes and the boundary branches, splitting the node admittance matrix, and generating independent admittance sub-matrices and equivalent boundary parameters; and distributing the independent admittance sub-matrices and the equivalent boundary parameters to different calculation units to realize parallel electromagnetic transient simulation. The application solves the problems of low parallel computing efficiency and insufficient boundary processing accuracy caused by unreasonable data splitting strategies in existing electromagnetic transient simulation of a power system, and improves the calculation efficiency and numerical accuracy of electromagnetic transient simulation of a large-scale power system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power simulation, and particularly relates to a data splitting processing method of an electromagnetic simulation model of a power system. BACKGROUND

[0002] A traditional electromagnetic transient simulation method adopts a centralized computing mode, models the entire power system as a unified mathematical model and solves it on a single computing platform. In the prior art, a numerical integration method based on a node admittance matrix is mainly adopted, a large-scale admittance matrix containing all power equipment and network connection relationships is constructed, an implicit or explicit numerical integration algorithm is used to solve a system of differential algebraic equations, so that time-domain response characteristics of each node voltage and branch current are obtained. This centralized simulation method has good calculation accuracy and numerical stability when processing small and medium-sized power systems, and has been widely applied in power system planning and operation analysis.

[0003] However, with the continuous expansion of the scale of the power system and the large-scale access of new energy, the existing centralized electromagnetic transient simulation method faces serious calculation bottlenecks and technical limitations. Firstly, the problem of calculation complexity, the dimension of the node admittance matrix of the large-scale power system is huge, the calculation complexity of the matrix solution increases exponentially with the system size, and a single computing unit is difficult to complete the simulation calculation within a reasonable time, which seriously restricts the real-time and practicality of the simulation. Secondly, the problem of memory requirement, a large-scale admittance matrix needs to occupy a large amount of memory space, which exceeds the memory capacity limit of a single computing platform, resulting in frequent memory exchange operations in the simulation process, which further reduces the calculation efficiency. In addition, the centralized computing mode lacks scalability, when the system size further increases, the problem of insufficient computing capacity cannot be solved by simply increasing the computing resources.

[0004] Based on the above analysis, the fundamental problem of the prior art is the lack of an effective data splitting method to realize parallel computation of electromagnetic transient simulation. The traditional method models and calculates the power system as an indivisible whole, fails to fully utilize the partition characteristics and local coupling characteristics of the power system itself, and leads to inefficient allocation and utilization of computing resources. Further, the existing network partition method is mainly based on static division based on geographical location or topological connection relationship, lacks in-depth consideration of the electrical characteristics of the power system, and the division result often leads to strong electrical coupling between subsystems, increasing the complexity of boundary processing and the frequency of data exchange. At the same time, the existing method lacks an accurate equivalent model for the boundary of the subsystem, and the simple boundary condition setting easily introduces numerical errors and calculation instability, affecting the accuracy and reliability of parallel simulation. Therefore, it is urgent to develop an intelligent data splitting method based on electrical characteristics, to realize reasonable division of the power system through accurate electrical distance calculation and clustering analysis, and to ensure electrical consistency between subsystems by using high-precision boundary equivalent processing technology. SUMMARY

[0005] The present application provides a data splitting method for a power system electromagnetic simulation model, which solves the problem of low parallel computation efficiency and insufficient boundary processing accuracy caused by unreasonable data splitting strategy in existing power system electromagnetic transient simulation, and improves the calculation efficiency and numerical accuracy of large-scale power system electromagnetic transient simulation.

[0006] The present application provides a data splitting method for a power system electromagnetic simulation model, which includes:

[0007] S1 step: Construct the node admittance matrix of the power system, and perform decomposition calculation on the node admittance matrix by using eigenvalue decomposition algorithm to form an electrical distance matrix;

[0008] S2 step: Based on the electrical distance matrix, execute K-means clustering algorithm to divide the nodes of the power system into multiple electrically coupled subsystems;

[0009] S3 step: Identify the boundary connection relationship between the multiple electrically coupled subsystems, and determine the boundary nodes and boundary branches of each electrically coupled subsystem;

[0010] S4 step: Perform Thevenin equivalent processing on the boundary nodes and boundary branches of each electrically coupled subsystem and split the node admittance matrix to generate independent admittance sub-matrix and equivalent boundary parameters of each electrically coupled subsystem;

[0011] S5 step: distributing the independent admittance sub-matrix of each electrical coupling subsystem and the equivalent boundary parameter to different calculation units, each calculation unit realizing parallel electromagnetic transient simulation based on boundary data exchange.

[0012] In the technical scheme provided in the application, the scheme of constructing a power system node admittance matrix and forming an electrical distance matrix by using an eigenvalue decomposition algorithm fundamentally changes the traditional static division mode based on geographical position or topological connection, and instead intelligently groups based on the inherent electrical characteristics of the power system. The eigenvalue decomposition algorithm can extract the inherent mathematical characteristics of the node admittance matrix, and through the calculation of eigenvalues and eigenvectors, it reveals the deep electrical coupling relationship of the power system network. The formation of the electrical distance matrix makes the similarity measurement between nodes more accurate and physically meaningful. The technical feature of performing the K-means clustering algorithm based on the electrical distance matrix further optimizes the rationality of node grouping. The K-means algorithm takes electrical distance as the similarity standard, ensuring that the nodes allocated to the same electrical coupling subsystem have high similarity in electrical characteristics, thereby significantly reducing the electrical coupling strength between subsystems and the boundary data exchange requirement. The technical feature of identifying the boundary connection relationship between multiple electrical coupling subsystems and determining the boundary nodes and boundary branches lays a data foundation for subsequent accurate boundary processing. Through the systematic boundary identification process, the interface position and connection mode of each subsystem can be accurately located. The technical feature of Thevenin equivalent processing of the boundary nodes and the boundary branches is the core innovation point of the application. The Thevenin equivalent processing simplifies the complex cross-subsystem electrical connection into a combination of equivalent voltage sources and equivalent impedances, which not only maintains the electrical accuracy at the boundary but also realizes the relative independence of the subsystems, creating the necessary conditions for parallel computing.

[0013] The eigenvalue decomposition algorithm can effectively identify the oscillation mode and electrical structure characteristics of the system in the application of power system, and its mathematical characteristics are highly consistent with the physical characteristics of the power system, so that the electrical distance calculation based on eigenvalue has clear physical meaning and engineering practical value. The K-means clustering algorithm shows good convergence and stability when dealing with the node grouping problem of the power system, and its iterative optimization mechanism can automatically adjust the clustering boundary to adapt to the complex network structure of the power system, avoiding the subjectivity and limitations of manual division. The Thevenin equivalent processing technology has a mature theoretical basis and a wide engineering application background in the boundary modeling of the power system, which can simplify the boundary condition setting and reduce the calculation dependency between subsystems while ensuring numerical accuracy. The technical features of distributing the independent admittance sub-matrix of each electrically coupled subsystem and the equivalent boundary parameters to different computing units realize true parallel processing, and the parallel electromagnetic transient simulation mechanism based on boundary data exchange of each computing unit not only ensures the global consistency of the calculation results but also fully utilizes the efficiency advantage of parallel computing, thereby significantly improving the calculation speed and resource utilization efficiency in large-scale power system simulation applications, solving the calculation bottleneck and scalability limitation problems faced by the traditional centralized simulation method in dealing with complex power networks. BRIEF DESCRIPTION OF DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor based on these drawings.

[0015] Figure 1 An embodiment schematic diagram of the data splitting processing method of the power system electromagnetic simulation model in the embodiment of the present application;

[0016] Figure 2 An embodiment schematic diagram of the calculation performance comparison between the traditional centralized method and the parallel method of the present application. DETAILED DESCRIPTION

[0017] The embodiments of the present application provide a data splitting processing method of an electromagnetic simulation model of a power system. The terms "first", "second", "third", "fourth" and the like (if any) in the specification and claims of the present application and the above-described drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" or "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to only those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to the process, method, product or device.

[0018] For ease of understanding, the specific processes of the embodiments of the present application are described below, please refer to Figure 1 One embodiment of the data splitting processing method of the electromagnetic simulation model of the power system in the embodiments of the present application includes:

[0019] S1 step: constructing a node admittance matrix of the power system, decomposing and calculating the node admittance matrix by using an eigenvalue decomposition algorithm to form an electrical distance matrix;

[0020] S2 step: executing a K-means clustering algorithm based on the electrical distance matrix to divide the nodes of the power system into a plurality of electrically coupled subsystems;

[0021] S3 step: identifying the boundary connection relationship between the plurality of electrically coupled subsystems, and determining the boundary nodes and boundary branches of each electrically coupled subsystem;

[0022] S4 step: performing Thevenin equivalent processing on the boundary nodes and boundary branches of each electrically coupled subsystem and splitting the node admittance matrix to generate an independent admittance sub-matrix and equivalent boundary parameters of each electrically coupled subsystem;

[0023] S5 step: distributing the independent admittance sub-matrix and equivalent boundary parameters of each electrically coupled subsystem to different computing units, and each computing unit realizes parallel electromagnetic transient simulation based on boundary data exchange.

[0024] It can be understood that the execution subject of the present application can be a data splitting processing system of an electromagnetic simulation model of a power system, and can also be a terminal or a server, and the specific embodiments are not limited herein. The embodiments of the present application take a server as an execution subject for example.

[0025] Specifically, a node admittance matrix is constructed, where the node admittance matrix is a mathematical representation describing the electrical connection relationship between nodes in the power system, and the matrix elements reflect the admittance values between nodes. The electrical parameters of various devices in the power system, including generators, transformers, transmission lines, etc., are collected, including resistance, reactance, conductance, etc. Then, according to Kirchhoff's current law and voltage law, a node voltage equation set is established, forming an n x n dimensional node admittance matrix, where n is the total number of system nodes. The eigenvalue decomposition algorithm decomposes the admittance matrix into the product form of the eigenvalue diagonal matrix and the eigenvector matrix. The eigenvalue reflects the intrinsic characteristics of the matrix, and the eigenvector represents the corresponding direction information. The electrical distance between any two nodes is calculated through the eigenvalue and eigenvector. The electrical distance comprehensively considers the electrical coupling strength and network topology relationship between nodes, and the smaller the value, the closer the electrical connection between the two nodes. The electrical distance of all node pairs is organized in row and column order to form an electrical distance matrix, which lays the data foundation for subsequent clustering analysis.

[0026] The K-means clustering algorithm is executed based on the electrical distance matrix. This algorithm is an unsupervised learning method that assigns data points to the nearest cluster center through iterative optimization. First, the number of clusters K is determined according to the size of the power system and the constraints of computing resources, generally set to the logarithm value of the number of system nodes. The K-means algorithm uses the electrical distance value as the similarity measurement standard, randomly selects K cluster centers initially, then calculates the distance from each node to each cluster center, and assigns the node to the nearest cluster center. Then the centroid position of each cluster is recalculated, the cluster center coordinates are updated, and the distance calculation and node assignment process is repeated until the cluster center position is stable or the maximum iteration number is reached. Finally, multiple electrically coupled subsystems are obtained, each containing a node set with similar electrical characteristics.

[0027] The boundary connection relationship between multiple electrically coupled subsystems is identified by analyzing the topological connection information of the original power system to check whether there is a direct electrical connection between the subsystems. The boundary node refers to a node that is connected to nodes of different subsystems, and the boundary branch refers to a transmission line or transformer device that connects nodes of different subsystems. All node connection relationships are traversed, and when a node is found to have a direct electrical connection with nodes belonging to different subsystems, the node is marked as a boundary node. Similarly, all branches are traversed, and when the nodes at both ends of a branch belong to different subsystems, the branch is marked as a boundary branch. The boundary nodes and boundary branches are paired and combined to form a boundary information list for each subsystem.

[0028] The boundary nodes and boundary branches of each electrically coupled subsystem are processed by the Thevenin equivalence, and the Thevenin theorem equivalently processes the complex linear circuit as a combination of a simple voltage source and a series impedance. For each boundary node, the equivalent voltage source and the equivalent impedance in the adjacent subsystem are calculated, the equivalent voltage source is equal to the voltage value of the node in the open circuit state, and the equivalent impedance is equal to the input impedance of the adjacent subsystem from the node. At the same time, the original node admittance matrix is processed by block according to the node attribution relationship of each subsystem, and each subsystem corresponds to an independent admittance sub-matrix. By rearranging the order of the matrix rows and columns, the nodes of the same subsystem are arranged continuously in the matrix, and then the corresponding sub-matrix block is extracted according to the subsystem boundary. The admittance sub-matrix of each subsystem and the equivalent boundary parameter are encapsulated and combined to form an independent data package.

[0029] The independent admittance sub-matrix and the equivalent boundary parameter of each electrically coupled subsystem are distributed to different calculation units for parallel processing. First, the calculation complexity of each subsystem is evaluated, considering factors such as the size of the subsystem, the type of components, the calculation accuracy requirement, and the load weight value of each subsystem is calculated. According to the load weight value and the number of available calculation units, a resource allocation strategy is formulated to distribute the calculation load relatively evenly to each calculation unit. Each calculation unit receives the allocated subsystem data package, and based on the admittance sub-matrix, an electromagnetic transient simulation model of the subsystem is established. A boundary data exchange mechanism is established between each calculation unit to exchange voltage and current information of the boundary nodes at the end of each simulation time step, ensuring the consistency and accuracy of the simulation results of each subsystem.

[0030] In a specific embodiment, S1 step, comprising:

[0031] Obtaining impedance parameters and branch admittance parameters of each node in the power system, and constructing a node admittance matrix;

[0032] Performing eigenvalue decomposition processing on the node admittance matrix to obtain an eigenvalue diagonal matrix and an eigenvector matrix;

[0033] Based on the eigenvalue diagonal matrix and the eigenvector matrix, the electrical distance value between any two nodes is calculated to obtain the electrical distance between the nodes;

[0034] Arranging and combining the electrical distance between the nodes according to the node serial number to obtain an electrical distance matrix.

[0035] Specifically, impedance parameters of each node in the power system and branch admittance parameters are acquired to construct a node admittance matrix, the impedance parameters refer to the degree of hindering current flow of the power system devices such as generators, transformers, transmission lines and the like, including resistance components and reactance components, while the branch admittance parameters are the inverse of the impedance parameters, indicating the ease of current flow through the branch. The node admittance matrix is a mathematical model describing the network characteristics of the power system, the diagonal elements of the matrix represent the self-admittance of each node, that is, the sum of the admittance of all branches directly connected to the node, and the non-diagonal elements represent the mutual admittance between nodes, which is a negative branch admittance value when there is a direct connection between two nodes, and zero when there is no direct connection. In the construction process, all nodes and branches of the power system need to be traversed, and the admittance value of each branch is filled into the corresponding position of the matrix according to the node position connected by it, forming a complete matrix reflecting the topology and electrical characteristics of the entire power network.

[0036] Eigenvalue decomposition of the node admittance matrix is a key mathematical operation for data splitting, and the eigenvalue decomposition algorithm decomposes the node admittance matrix into the product form of the eigenvalue diagonal matrix and the eigenvector matrix. The eigenvalue reflects the inherent mathematical properties of the matrix, and each eigenvalue corresponds to the scaling factor of the matrix in a certain direction, and the eigenvector matrix contains the direction vectors corresponding to each eigenvalue, which form the basis of matrix transformation. The decomposition process first calculates the characteristic polynomial of the node admittance matrix, solves the characteristic equation to obtain all eigenvalues, then solves the corresponding eigenvector for each eigenvalue, arranges the eigenvalues in the diagonal to form the eigenvalue diagonal matrix, and arranges the eigenvectors in columns to form the eigenvector matrix. This decomposition reveals the internal structural characteristics of the power system network, laying a mathematical foundation for subsequent electrical distance calculation.

[0037] The calculation of the electrical distance between any two nodes based on the eigenvalue diagonal matrix and the eigenvector matrix is the core innovation of the present application, which comprehensively considers the electrical coupling strength and network topology relationship between nodes, and is different from the traditional geographical distance or impedance distance. In the calculation process, the row vectors corresponding to the two nodes in the eigenvector matrix are taken, the difference between the two row vectors in each eigenvector dimension is calculated, then the square of the difference is multiplied by the weight of the corresponding eigenvalue and summed, and finally the square root is taken to obtain the electrical distance value. The weight of the eigenvalue reflects the importance of this dimension to the electrical relationship between nodes, the larger the eigenvalue, the more significant the influence of this dimension, and through weighted processing, the actual electrical contact strength between nodes can be more accurately reflected. The smaller the electrical distance value, the more similar the two nodes are in electrical characteristics, and the more intense the mutual influence in electromagnetic transient simulation.

[0038] The electrical distance matrix is obtained by arranging and combining the electrical distances between nodes according to the node numbers, and is the final link of data organization. The electrical distance matrix is a symmetric matrix, and the element in the ith row and jth column represents the electrical distance value between node i and node j. The arrangement and combination process needs to calculate the electrical distance of any two node pairs in the system. For a system containing n nodes, the distance values of n times n minus one divided by two node pairs need to be calculated, and then the distance data is organized in the form of a matrix. The diagonal elements of the matrix are zero because the electrical distance of any node to itself is zero. The non-diagonal elements are filled in the corresponding positions according to the node number order to form a complete electrical distance matrix. The matrix not only retains the connection information of the original network, but also integrates the electrical characteristic information.

[0039] In a specific embodiment, the S2 step comprises:

[0040] Based on the electrical distance matrix, the number of clusters K is determined, and the clustering parameter of the K-means clustering algorithm is set;

[0041] The electrical distance values in the electrical distance matrix are input as similarity measurement standards into the K-means clustering algorithm to select the initial cluster center coordinates;

[0042] According to the initial cluster center coordinates and the clustering parameter, the distance calculation and classification allocation processing of each node of the power system are performed to obtain the node clustering grouping result;

[0043] The node clustering grouping result is iteratively optimized to update the cluster center position, and a plurality of electrical coupling subsystems are obtained.

[0044] Specifically, the K value represents how many electrical coupling subsystems the power system nodes are divided into, and its determination needs to comprehensively consider the scale of the power system, the available computing resources and the parallel processing capability. The clustering parameters include the maximum number of iterations, the convergence precision threshold, the initialization method and other key indicators to control the algorithm execution. The maximum number of iterations limits the upper limit of the running time of the algorithm, the convergence precision threshold determines the judgment standard of the stability of the cluster center position, and the initialization method determines the selection strategy of the initial cluster center. The determination process of the K value first counts the total number of nodes of the power system, and then calculates the reasonable grouping number according to the appropriate node scale range of each subsystem, while considering the number limitation of the computing unit to ensure the relative balance of the computing load of each subsystem. The setting of the clustering parameter needs to balance the calculation accuracy and the running efficiency. Too high convergence precision leads to too long algorithm running time, and too low convergence precision affects the clustering quality and data splitting effect.

[0045] The electrical distance value in the electrical distance matrix is input into the K-means clustering algorithm as a similarity measure for initial cluster center selection, which is the core operation of data processing. K-means clustering algorithm is a distance-based unsupervised learning method that realizes data grouping by minimizing the sum of squares of distances from each data point to its cluster center. The electrical distance value replaces the traditional Euclidean distance as a similarity measure, fully considering the electrical coupling relationship between nodes in the power system rather than simply the spatial position relationship. The initial cluster center selection uses random selection method or K-means++ improvement method. The random selection method randomly selects K nodes from all nodes as initial centers, and the K-means++ method selects initial centers according to distance weights to improve the convergence speed of the algorithm. The initial cluster center coordinates are actually the row vectors of the corresponding nodes in the electrical distance matrix, and each coordinate component represents the electrical distance relationship between the node and other nodes in the system. The selection of cluster centers directly affects the convergence effect of the subsequent iteration process and the quality of the final grouping result.

[0046] The distance calculation and classification assignment process of each node in the power system according to the initial cluster center coordinates and clustering parameters is the execution phase of the clustering algorithm. The distance calculation process traverses each node in the system and calculates the electrical distance value between the node and each cluster center. The classification assignment process assigns each node to the subsystem represented by the nearest cluster center, forming the initial node grouping relationship. Distance calculation uses the precomputed results stored in the electrical distance matrix to avoid repeated calculation and improve processing efficiency. For node i and cluster center j, the corresponding distance value is directly obtained from the i-th row and j-th column position of the electrical distance matrix. The classification assignment process creates a mapping relationship table of nodes to subsystems, recording the subsystem number and corresponding distance value of each node. The node clustering grouping result contains the node member list of each subsystem and the distance information of each node to its cluster center. These data are the basis for subsequent iteration optimization and boundary identification.

[0047] The iterative optimization of the node clustering grouping result to update the cluster center position is the core iteration link of the algorithm, and the iterative optimization process repeatedly performs the two steps of cluster center update and node reassignment until the algorithm converges. The cluster center update calculates the barycenter position of all nodes in each subsystem, and the barycenter position is the geometric center of each node in the electrical distance space in the subsystem. The new cluster center coordinates are obtained by performing arithmetic averaging on the electrical distance vectors of all nodes in the subsystem. The node reassignment step reassigns the nodes to the nearest new cluster center based on the updated cluster center. The algorithm converges to judge the change amount of the cluster center position in the two consecutive iterations, and when the change amount is less than the preset convergence accuracy threshold, it is considered that the algorithm converges, and the final multiple electrical coupling subsystems are output. The electrical coupling subsystem is a node set with similar electrical characteristics, and the nodes in the same subsystem are close to each other in electrical distance. The nodes between different subsystems are relatively far away in electrical distance. This grouping method is beneficial to reduce the electrical coupling strength and data exchange frequency between subsystems.

[0048] In a specific embodiment, the S3 step includes:

[0049] Based on the node membership relationship of the multiple electrical coupling subsystems, a topological connection analysis is performed to obtain a connection topological structure between the subsystems;

[0050] According to the connection topological structure between the subsystems, a connection node that spans different electrical coupling subsystems is identified to obtain a boundary node list;

[0051] Branch connection relationship retrieval is performed on the nodes in the boundary node list to obtain boundary branches that connect different electrical coupling subsystems;

[0052] The boundary node list and the boundary branches that connect different electrical coupling subsystems are paired and combined to obtain boundary nodes and boundary branches of each electrical coupling subsystem.

[0053] Specifically, topology connection analysis needs to compare the complete network topology of the original power system with the node grouping result after clustering analysis, and identify which nodes originally directly connected are allocated to different subsystems. The node attribution relationship is the mapping table of nodes to subsystems generated by the pre-order K-means clustering algorithm, which records the subsystem number to which each node belongs, while the original network topology contains all the direct connection relationships between nodes, usually stored in the form of an adjacency matrix or a connection list. The topology connection analysis process traverses each connection relationship in the original network, checks whether the nodes at both ends of the connection belong to the same subsystem, and when it is found that the nodes at both ends of the connection belong to different subsystems, the connection constitutes a cross-connection between subsystems. The inter-subsystem connection topology structure is a data structure that describes the direct electrical connection relationship between subsystems, usually represented by a subsystem-level adjacency matrix, and the matrix element value represents the number or strength of the connection between the two subsystems. This structure abstracts the overall connection relationship between subsystems and screens out the specific details of the internal nodes.

[0054] According to the inter-subsystem connection topology structure, the connection nodes that cross different electrically coupled subsystems are identified as the core operation of boundary node identification. Connection nodes refer to nodes that have direct connection relationships with multiple nodes within a subsystem. The boundary node identification algorithm traverses each cross-connection between subsystems, extracts the node information at both ends of the connection, and these nodes naturally have the characteristics of cross-subsystem connection and are key nodes for information transmission and energy exchange between electrically coupled subsystems. The identification process creates a node connection statistics table to record the number of connections and connection objects of each node with other subsystem nodes. When the number of cross-subsystem connections of a node is greater than zero, the node is marked as a boundary node. The boundary node list is grouped and sorted according to the subsystem attribution, and each subsystem's boundary node list contains all nodes within the subsystem that have cross-system connection characteristics. These nodes play a key interface role in the subsequent Thevenin equivalent processing. The accuracy of boundary node identification directly affects the accuracy of data exchange between subsystems and the consistency of simulation results.

[0055] The branch connection relationship retrieval of the nodes in the boundary node list is a detailed analysis process of the boundary branch identification. The branch connection relationship retrieval needs to find all the connection branches of each boundary node in the original power system, and filter out the branches connected to the nodes of other subsystems. The branch is an electrical equipment connecting two nodes in the power system, including transmission lines, transformers, reactors, etc. Each branch has specific electrical parameters such as resistance, reactance, and admittance. The retrieval process queries all the branch information connected to each boundary node from the branch database of the power system, and then checks whether the other end node of each branch belongs to a different subsystem. When the two nodes connected by a branch belong to different subsystems, the branch is identified as a boundary branch. The boundary branch connecting different electrical coupling subsystems is the physical carrier for realizing the electrical connection between the subsystems, and carries the power transmission and voltage transfer functions between the subsystems. The electrical parameters of these branches directly affect the coupling strength between the subsystems and the setting of the boundary conditions for simulation calculation.

[0056] The pairing and combination processing of the boundary node list and the boundary branches connecting different electrical coupling subsystems is the final arrangement step for establishing complete boundary information. The pairing and combination processing needs to establish the correspondence between the boundary nodes and the boundary branches, to ensure that each boundary branch can be accurately associated with the boundary node connected thereto. The pairing process creates a boundary information association table, which records the list of boundary branches connected to each boundary node and the pair of boundary nodes connected by each boundary branch. This bidirectional association relationship facilitates subsequent Thevenin equivalent calculation and boundary data exchange processing. The combination processing also needs to classify and arrange the boundary information according to the subsystem attribution. The boundary nodes and boundary branches of each subsystem contain the interface information between the subsystem and all other subsystems, including the internal number of the boundary node of the subsystem, the target subsystem number of the cross-system connection, the electrical parameters of the connection branch, etc. The boundary nodes and boundary branches of each electrical coupling subsystem constitute a complete description of the subsystem boundary. These information is the direct input data for subsequent Thevenin equivalent processing and matrix splitting.

[0057] In a specific embodiment, the S4 step comprises:

[0058] Calculating Thevenin equivalent voltage sources and equivalent impedance parameters based on the boundary nodes and boundary branches of each electrical coupling subsystem to obtain boundary equivalent parameters;

[0059] Performing equivalent replacement processing on the boundary connection of each electrical coupling subsystem according to the boundary equivalent parameters to obtain an independent subsystem model;

[0060] Performing matrix block splitting processing on the node admittance matrix according to the nodes of each electrical coupling subsystem to obtain the corresponding admittance sub-matrix of each electrical coupling subsystem;

[0061] The data combination and packaging are performed based on the boundary equivalent parameters and the admittance sub-matrix to obtain the independent admittance sub-matrix and equivalent boundary parameters of each electrical coupling subsystem.

[0062] Specifically, the Thevenin theorem equivalently represents a complex linear circuit as a combination of a simple ideal voltage source and a series impedance, so that the electrical connection relationship between the subsystems is simplified. The Thevenin equivalent voltage source refers to the open-circuit voltage value generated by other subsystems at a boundary node when viewed from the outside. The calculation process needs to disconnect the boundary node from the subsystem to which it belongs, and then calculate the comprehensive voltage effect generated by all other subsystems at the node through the boundary branch. The equivalent impedance parameter refers to the equivalent input impedance when viewed from the boundary node into other subsystems. When calculating, all voltage sources in other subsystems are short-circuited, current sources are open-circuited, and only impedance elements are retained, and then the total equivalent impedance value is calculated from the boundary node. The calculation of the boundary equivalent parameters needs to be performed for each boundary node respectively, considering the electrical connection relationship between the node and each of the other subsystems, and the corresponding equivalent voltage source and equivalent impedance values are obtained by solving the node voltage equation and the branch current equation. These parameters completely describe the electrical boundary conditions at the boundary node.

[0063] The equivalent substitution of the boundary connection of each electrical coupling subsystem according to the boundary equivalent parameters is the core operation of the subsystem independence, and the equivalent substitution replaces the originally complex cross-subsystem electrical connection relationship with a simple Thevenin equivalent circuit. During the substitution process, the external connection of each boundary node is replaced by a series circuit of an equivalent voltage source and an equivalent impedance. The value of the equivalent voltage source is equal to the Thevenin equivalent voltage calculated above, and the value of the equivalent impedance is equal to the Thevenin equivalent impedance calculated above. The independent subsystem model refers to a subsystem that no longer directly includes the specific electrical elements of other subsystems after the equivalent substitution, but interacts with the external environment through equivalent boundary conditions. This independent processing makes the electrical model of each subsystem relatively independent, and the electrical calculation inside each subsystem no longer directly depends on the detailed parameters and operating state of other subsystems, but only reflects the comprehensive influence effect of the external subsystem through the boundary equivalent parameters, thereby creating necessary conditions for parallel computing.

[0064] The matrix block splitting of the node admittance matrix according to the node grouping of each electrical coupling subsystem is a specific implementation step of the data physical separation. The matrix block splitting needs to rearrange and split the original admittance matrix according to the node attribution relationship obtained by the cluster analysis. The splitting process first reorders the rows and columns of the admittance matrix, so that the nodes belonging to the same subsystem are arranged continuously in the matrix, forming a reordered admittance matrix with a block diagonal structure. Then, the reordered matrix is divided into several independent sub-matrix blocks according to the node boundary positions of each subsystem, and each sub-matrix block corresponds to the internal admittance relationship of an electrical coupling subsystem. The admittance sub-matrix corresponding to each electrical coupling subsystem contains the admittance connection relationship between all nodes in the subsystem, but does not contain the direct admittance connection with the nodes of other subsystems. These admittance connections across the subsystems are replaced by the aforementioned Thevenin equivalence processing. The dimension of the admittance sub-matrix is equal to the number of nodes in the corresponding subsystem, and the matrix elements reflect the electrical network characteristics within the subsystem, which is the basic mathematical model for independent electromagnetic transient simulation calculation of each subsystem.

[0065] The data combination and packaging based on the boundary equivalent parameters and the admittance sub-matrix are the final integration steps to form a complete subsystem data package. The data combination and packaging integrate the internal admittance relationship and external boundary conditions of each subsystem into an independent data unit. The packaging process creates a standardized subsystem data structure, which includes the admittance sub-matrix, boundary node information, equivalent voltage source parameters, equivalent impedance parameters, subsystem identifier and other complete information. The independent admittance sub-matrix of each electrical coupling subsystem can independently describe the electrical behavior characteristics of the subsystem under given boundary conditions after being modified by the boundary equivalent parameters. The equivalent boundary parameters, as the interface information between the subsystem and the external environment, contain the Thevenin equivalent voltage, equivalent impedance and adjacent subsystem identifier of the boundary node. These parameters are used for data exchange and boundary condition updating between subsystems in parallel simulation. The data packaging also includes the mapping relationship between the internal node number and the global node number, the communication protocol definition of the boundary data exchange between subsystems and other auxiliary information, which ensures the integrity and independence of each subsystem data package.

[0066] In a specific embodiment, the process of performing the step of matrix block splitting of the node admittance matrix according to the node grouping of each electrical coupling subsystem can specifically include the following steps:

[0067] Based on the node attribution relationship of each electrical coupling subsystem, the rows and columns of the node admittance matrix are indexed and reordered to obtain a reordered node admittance matrix;

[0068] According to the number of nodes of each electrical coupling subsystem, the matrix block boundary positions are determined to obtain the matrix block index coordinates;

[0069] The reordered node admittance matrix is subjected to a sub-matrix extraction operation according to the matrix block index coordinates to obtain original admittance sub-matrices of each electrical coupling subsystem;

[0070] The original admittance sub-matrices of each electrical coupling subsystem are subjected to a matrix integrity inspection and a data format standardization processing to obtain admittance sub-matrices corresponding to each electrical coupling subsystem.

[0071] Specifically, the row and column index sorting of the node admittance matrix based on the node attribution relationship of each electrical coupling subsystem is a basic operation of matrix reorganization, which requires renumbering and rearranging the rows and columns of the original admittance matrix according to the node grouping results obtained by K-means clustering analysis. The node attribution relationship is a mapping table that records which electrical coupling subsystem each node belongs to and its local number in the subsystem. This mapping relationship is the basis and guiding principle for matrix reordering. The index sorting process first performs primary sorting of the nodes according to the subsystem number, so that the nodes belonging to the same subsystem are arranged continuously in the new number sequence. Then, secondary sorting is performed within each subsystem according to the original node number or a specific rule to ensure the consistency and traceability of the node order within the subsystem. The reordered node admittance matrix refers to the admittance matrix reorganized according to the new node number sequence, in which the nodes of the same subsystem originally scattered in various positions of the matrix are now concentrated in the continuous block area of the matrix. This reordering operation changes the position distribution of the matrix elements, but does not change the numerical value of the matrix elements. The mathematical properties and physical meanings of the matrix remain unchanged, and only the organization form of the data is adjusted.

[0072] Determining the matrix block boundary positions according to the number of nodes in each electrical coupling subsystem is a key step in developing the splitting strategy. The matrix block boundary positions refer to the starting row and column coordinates and the ending row and column coordinates of each sub-matrix in the reordered admittance matrix. The determination process of the boundary positions requires counting the number of nodes included in each electrical coupling subsystem, which directly determines the dimension size of the corresponding sub-matrix and its distribution position in the overall matrix. The calculation of the matrix block index coordinates uses the cumulative summation method. The starting coordinates of the first subsystem are the first row and first column of the matrix, and the ending coordinates are the row and column positions corresponding to the number of nodes in the first subsystem. The starting coordinates of the second subsystem are immediately after the ending position of the first subsystem, and the ending coordinates are the positions corresponding to the sum of the number of nodes in the first two subsystems. This process is repeated until the boundary coordinates of all subsystems are determined. The block index coordinates not only contain the row and column range information of each sub-matrix corresponding to the subsystem, but also contain auxiliary information such as the subsystem identifier, the number of nodes, and the dimension of the sub-matrix. These coordinate information is the direct input parameter and operation instruction for the subsequent sub-matrix extraction operation.

[0073] The sub-matrix extraction operation according to the block index coordinates extracts the independent matrix block corresponding to each electrical coupling subsystem from the reordered matrix. The extraction operation uses the matrix slicing method. For each subsystem, the corresponding starting row and column coordinates and the ending row and column coordinates are used to cut the corresponding matrix region from the reordered admittance matrix to form an independent sub-matrix data structure. The original admittance sub-matrix of each electrical coupling subsystem refers to the sub-matrix directly extracted from the reordered matrix without further processing. The sub-matrix retains all numerical information and structural characteristics of the corresponding region in the original admittance matrix. The dimension of the original admittance sub-matrix is equal to the number of nodes of the corresponding subsystem, and the matrix elements reflect the direct admittance connection relationship between the nodes in the subsystem. However, since only the nodes within the subsystem are considered during extraction, the admittance connection information across the subsystems is temporarily ignored at this stage and needs to be supplemented and corrected through subsequent boundary processing steps.

[0074] The matrix integrity test and data format standardization processing of the original admittance sub-matrix of each electrical coupling subsystem are necessary links for quality control and format unification. The matrix integrity test includes matrix dimension verification, numerical range check, symmetry verification, singularity judgment and other detection operations. The integrity test first verifies whether the number of rows and columns of the sub-matrix is equal to the number of nodes of the corresponding subsystem to ensure the correctness of the matrix dimension, and then checks whether the numerical value of the matrix element is within a reasonable range to avoid numerical overflow or the existence of abnormal values. The symmetry verification checks whether the admittance sub-matrix satisfies the properties of a symmetric matrix. Since the node admittance matrix should be symmetric under normal circumstances, asymmetry usually indicates errors in the data extraction process or problems with the original data. The data format standardization processing includes numerical precision unification, storage format specification, index number adjustment and other operations to ensure that each sub-matrix uses a consistent data representation method and storage structure. The admittance sub-matrix corresponding to each electrical coupling subsystem is the final sub-matrix after integrity test and format standardization.

[0075] In a specific embodiment, the process of determining the matrix block boundary position according to the number of nodes of each electrical coupling subsystem can specifically include the following steps:

[0076] Count the number of nodes contained in each electrical coupling subsystem to obtain the node count value of each electrical coupling subsystem;

[0077] Based on the node count value of each electrical coupling subsystem, perform an accumulation calculation to obtain the starting position index of each electrical coupling subsystem in the matrix;

[0078] The starting position index of each electrical coupling subsystem and the node count value are used to calculate the ending position index of each subsystem, so as to obtain the matrix boundary range of each electrical coupling subsystem;

[0079] The starting position index of each electrical coupling subsystem and the matrix boundary range are combined to obtain the matrix block index coordinates.

[0080] Specifically, the node quantity statistics need to traverse the node attribution table generated by the pre-order K-means clustering analysis, and count the number of nodes included in each electrical coupling subsystem one by one. The node count value refers to the total number of power system nodes included in each electrical coupling subsystem, including generator nodes, transformer nodes, bus nodes, load nodes and network nodes corresponding to various electrical equipment. The statistical process adopts the grouping counting method, creates a node counter for each subsystem, sets the initial value to zero, then traverses each record in the node attribution table, and adds one to the corresponding counter value according to the subsystem number to which the node belongs. After the traversal is completed, the value of each counter is the node count value of the corresponding subsystem. The node count value of each electrical coupling subsystem not only reflects the size of the subsystem, but also directly determines the dimension of the corresponding admittance submatrix, because the admittance submatrix is a square matrix, and its row number and column number are equal to the number of nodes in the subsystem. The accuracy of the node count value directly affects the correctness of the subsequent matrix block operation, and incorrect counting will cause dimension mismatch or data out-of-bound problems when extracting the submatrix.

[0081] The accumulation calculation based on the node count value of each electrical coupling subsystem is a key calculation link for determining the distribution position of the submatrix in the overall matrix. The accumulation calculation adopts the prefix sum algorithm to determine the starting position of the corresponding submatrix of each subsystem by gradually accumulating the node count values of each subsystem. The starting position index of each electrical coupling subsystem in the matrix refers to the starting row number and starting column number of the admittance submatrix corresponding to the subsystem in the reordered node admittance matrix. Since the admittance matrix is a symmetric square matrix, the starting row number and the starting column number are equal. In the calculation process of the starting position index, the starting position index of the first subsystem is one, indicating that it starts from the first row and the first column of the matrix, the starting position index of the second subsystem is equal to the node count value of the first subsystem plus one, the starting position index of the third subsystem is equal to the sum of the node count values of the first two subsystems plus one, and so on until the starting position indices of all subsystems are calculated. The accumulation calculation process is actually a prefix sum operation on the node count value sequence. The starting position of each subsystem is equal to the sum of the node count values of all previous subsystems plus one. This calculation method ensures that each subsystem occupies a continuous and non-overlapping area in the matrix.

[0082] The calculation of the end position index of each electrical coupling subsystem is a calculation step for completing the definition of the sub-matrix boundary, and the calculation of the end position index uses the method of adding the start position index to the node count value minus one. The matrix boundary range of each electrical coupling subsystem refers to the row and column range occupied by the admittance sub-matrix corresponding to the subsystem in the reordered node admittance matrix, which is defined by the start position index and the end position index. Attention should be paid to the continuity and integrity of the index during the calculation of the end position index. The end position of each subsystem should be equal to the start position of the next subsystem minus one, to ensure that the distribution of each subsystem in the matrix is continuous and has no overlap or omission. The matrix boundary range not only includes the start and end row and column index values, but also includes the dimension information of the sub-matrix, i.e. the number of rows and columns, which are direct input parameters for subsequent sub-matrix extraction operations. The calculation of the boundary range needs to verify its rationality, including checking whether the end position is greater than the start position, whether the boundaries of each subsystem are continuous, and whether the union of the boundary ranges of all subsystems covers the entire matrix.

[0083] The combination of the start position index of each electrical coupling subsystem and the matrix boundary range is a data integration step for forming a complete index coordinate system. The combination of the coordinate pairs integrates the scattered position information into structured index data. The matrix block index coordinate is a complete coordinate system that describes the position and range of each subsystem admittance sub-matrix in the overall matrix. Each subsystem corresponds to a set of coordinate parameters, including the start row index, the start column index, the end row index, the end column index, the number of rows of the sub-matrix, the number of columns of the sub-matrix, and other complete information. The coordinate pair combination process creates a standardized index data structure, which organizes and stores the position information of each subsystem in a unified format, facilitating direct invocation and use by subsequent sub-matrix extraction operations. The index coordinate also includes subsystem identifiers, node number mapping tables, boundary node markers, and other auxiliary information, which play an important role in sub-matrix extraction and subsequent boundary processing. The result of the coordinate combination is a complete index table containing the position information of all subsystems.

[0084] In a specific embodiment, the S5 step includes:

[0085] Based on the calculation complexity and node size of each electrical coupling subsystem, a load weight value of each electrical coupling subsystem is obtained;

[0086] According to the load weight value of each electrical coupling subsystem and the number of available computing units, a resource allocation calculation is performed to obtain a computing unit allocation scheme;

[0087] The independent admittance sub-matrix and equivalent boundary parameters of each electrical coupling subsystem are distributed to the corresponding computing unit according to the computing unit allocation scheme, to obtain a simulation data package for each computing unit;

[0088] Based on the simulation data packets of each computing unit, parallel electromagnetic transient calculation is started and a boundary data exchange communication mechanism is established to obtain electromagnetic transient simulation results.

[0089] Specifically, the calculation load evaluation needs to comprehensively consider multiple load influencing factors of each subsystem to determine the calculation demand thereof. The calculation complexity refers to the operation amount and time consumption required for electromagnetic transient simulation calculation of each electrically coupled subsystem, which is mainly determined by the types and quantities of power equipment contained in the subsystem. Different types of power equipment have different calculation complexity coefficients. The generator needs to solve complex electromechanical transient differential equations, and has a higher calculation complexity. The simple load node only needs to perform algebraic operation, and has a lower calculation complexity. The node scale refers to the total number of nodes contained in each electrically coupled subsystem, which directly affects the dimension of the admittance matrix and the calculation amount of matrix operation. The more the number of nodes, the larger the size of the equation group to be solved, and the heavier the calculation load of matrix operation. The load evaluation process first counts the number of different types of equipment in each subsystem, then calculates the equipment complexity load of each subsystem according to the calculation complexity coefficient of the equipment type, then calculates the matrix operation load according to the number of nodes, and finally combines the equipment complexity load and the matrix operation load to obtain the load weight value of each electrically coupled subsystem. The load weight value is a numerical index for describing the relative size of the calculation demand of each subsystem. The greater the weight value, the more calculation resources and processing time the subsystem needs.

[0090] Resource allocation calculation according to the load weight value of each electrically coupled subsystem and the number of available computing units is the core step of formulating a parallel calculation strategy. The resource allocation calculation needs to find the optimal balance point between load balancing and calculation efficiency. The number of available computing units refers to the total number of processor cores or computing nodes used for parallel electromagnetic transient simulation, which limits the maximum concurrency of parallel processing and the flexibility of resource allocation. The resource allocation calculation adopts a load balancing algorithm, the goal of which is to make the load weight of each computing unit as equal as possible, avoiding the situation that some computing units are overloaded while others are idle. The allocation algorithm first calculates the sum of the load weight values of all subsystems, then divides the ideal average load of a single computing unit by the number of available computing units, and then uses a greedy algorithm or a dynamic programming algorithm to allocate each subsystem to different computing units, so that the load weight value of each computing unit is close to the ideal average load. The computing unit allocation scheme is an allocation table describing the correspondence between each electrically coupled subsystem and the computing unit, recording which computing unit each subsystem is allocated to and which subsystem each computing unit is responsible for processing. The scheme directly guides the subsequent data distribution and task scheduling operations.

[0091] The specific execution step of data distribution is to distribute the independent admittance sub-matrix of each electrical coupling subsystem and the equivalent boundary parameter to the corresponding calculation unit according to the calculation unit allocation scheme. The data distribution process needs to accurately transmit the subsystem data packet generated by the split processing of the previous data to the specified calculation unit. The independent admittance sub-matrix is a mathematical description of the internal electrical connection relationship of each electrical coupling subsystem, which contains the admittance parameters and topological structure information between all nodes in the subsystem. The equivalent boundary parameter is the Thevenin equivalent circuit parameter describing the electrical connection between the subsystem and the external environment, including the equivalent voltage source and equivalent impedance value. The data distribution process determines the target calculation unit of each subsystem data packet according to the calculation unit allocation scheme, and then transmits the data packet to the corresponding calculation unit through network communication or shared memory. The simulation data packet of each calculation unit refers to the complete data set received by each calculation unit, which contains the admittance sub-matrix, boundary equivalent parameter, node number mapping, subsystem identification information and other complete data of all subsystems handled by the calculation unit. The simulation data packet also contains simulation parameter settings, time step, precision requirements, output format and other control information, which ensures that each calculation unit can independently perform electromagnetic transient simulation calculation.

[0092] The final execution phase of distributed simulation is to start parallel electromagnetic transient calculation based on the simulation data packet of each calculation unit and establish a boundary data exchange communication mechanism. Parallel electromagnetic transient calculation refers to the simultaneous electromagnetic transient simulation of each subsystem handled by each calculation unit. Electromagnetic transient simulation needs to solve the differential algebraic equation set of each subsystem, including the electromechanical transient equation of the generator, the electromagnetic transient equation of the transmission line, the algebraic equation of the load, etc. The calculation process uses numerical integration method to gradually advance the simulation time according to the preset time step. The boundary data exchange communication mechanism is a key technology to ensure the consistency of the simulation results of each subsystem. Since each subsystem is electrically connected through the boundary node, each calculation unit needs to exchange the voltage and current information of the boundary node at the end of each time step. The communication mechanism adopts a synchronous communication mode, where all calculation units wait for other units to complete after completing the calculation of the current time step, and then exchange boundary data uniformly to update the boundary conditions of each subsystem, and then start the calculation of the next time step. The electromagnetic transient simulation result is the complete system simulation data generated by the collaborative calculation of each calculation unit, including the voltage waveform of each node, the current waveform of each branch, the operating state of each device, etc.

[0093] Figure 2 The schematic diagram for performance comparison between the traditional centralized method and the parallel method of the present application in the embodiments of the present application. As shown in FIG. 1, the traditional centralized method is a single calculation unit, which needs to handle all subsystems and exchange boundary data between subsystems. The parallel method of the present application is a distributed calculation unit, which needs to handle only the subsystems allocated to it and exchange boundary data with other calculation units. Figure 2As shown, the test adopts IEEE standard test systems, including IEEE 118 nodes, IEEE 300 nodes and an extended large-scale test system, and the node scale is expanded from 100 to 8000. The ordinate adopts a logarithmic scale to represent the single-step electromagnetic transient simulation calculation time, and the time step is set to 50 microseconds, which meets the typical precision requirement of electromagnetic transient simulation.

[0094] The calculation complexity of the traditional centralized method (square mark solid line) is mainly limited by the solution of the node admittance matrix, and the time complexity increases by the second and a half power of the number of nodes, which is determined by the inherent calculation characteristics of the large-scale sparse matrix LU decomposition. When the system scale increases from 100 nodes to 8000 nodes, the calculation time of the traditional method increases from 0.05 seconds to about 2000 seconds, and the growth factor is 40000 times, which meets the "dimension disaster" phenomenon recognized in the field of power system simulation.

[0095] The parallel method of the application (circular mark dotted line) realizes performance breakthrough through the following technical links: first, the calculation complexity of the original third power is reduced to the second power based on the electrical distance calculation of the node admittance matrix eigenvalue decomposition; second, the calculation complexity of the K-means clustering algorithm is proportional to the number of subsystems, the number of nodes and the number of iterations, wherein the number of subsystems is usually set to 4 to 8, and the number of iterations is usually converged within 15 to 20 times; finally, the boundary node number involved in the Thevenin equivalent treatment usually accounts for 5% to 15% of the total number of nodes, and the calculation overhead of the boundary treatment is relatively small. Under the scale of 8000 nodes, the method of the application divides the system into 6 electrically coupled subsystems, each subsystem contains about 1300 nodes on average, and the solution time of the admittance submatrix of a single subsystem is about 120 seconds. The theoretical time of parallel calculation of 6 subsystems is 120 seconds. Considering the boundary data exchange overhead (about 8% to 12% of the total time) and the influence of load imbalance (the load balancing coefficient is about 1.15), the actual parallel calculation time is about 700 seconds, which is about 65% shorter than the 2000 seconds of the traditional method.

[0096] The performance data is verified based on the standard test environment, and the actual test is carried out by using a 6-core processor and a 32GB memory configuration. The measured data verifies that the application significantly improves the calculation efficiency of large-scale power system simulation through intelligent data splitting processing under the premise of maintaining the accuracy of electromagnetic transient level simulation, and solves the problem of low parallel calculation efficiency caused by unreasonable data splitting strategy in the prior art.

[0097] The above examples are only used to illustrate the technical solutions of the present application, and are not intended to limit the present application; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that the technical solutions recorded in the foregoing examples can be modified, or some technical features can be replaced by equivalent features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A data splitting processing method of an electromagnetic simulation model of a power system, characterized by, The method comprises: S1 step: constructing the node admittance matrix of the power system, decomposing and calculating the node admittance matrix by using an eigenvalue decomposition algorithm to form an electrical distance matrix, comprising: obtaining impedance parameters and branch admittance parameters of each node in the power system, and constructing a node admittance matrix; performing eigenvalue decomposition processing on the node admittance matrix to obtain an eigenvalue diagonal matrix and an eigenvector matrix; calculating the electrical distance value between any two nodes based on the eigenvalue diagonal matrix and the eigenvector matrix to obtain the electrical distance between nodes; arranging and combining the electrical distance between nodes according to the node serial number to obtain the electrical distance matrix; S2 step: performing a K-means clustering algorithm based on the electrical distance matrix to divide the nodes of the power system into a plurality of electrically coupled subsystems; S3 step: identifying the boundary connection relationship between the plurality of electrically coupled subsystems, and determining the boundary nodes and boundary branches of each electrically coupled subsystem; S4 step: performing Thevenin equivalent processing on the boundary nodes and boundary branches of each electrically coupled subsystem and splitting the node admittance matrix to generate an independent admittance sub-matrix and equivalent boundary parameters of each electrically coupled subsystem; S5 step: distributing the independent admittance sub-matrix and equivalent boundary parameters of each electrically coupled subsystem to different calculation units, and each calculation unit realizes parallel electromagnetic transient simulation based on boundary data exchange, comprising: performing calculation load evaluation based on the calculation complexity and node scale of each electrically coupled subsystem to obtain the load weight value of each electrically coupled subsystem; performing resource allocation calculation according to the load weight value of each electrically coupled subsystem and the number of available calculation units to obtain a calculation unit distribution scheme; distributing the independent admittance sub-matrix and equivalent boundary parameters of each electrically coupled subsystem to the corresponding calculation unit according to the calculation unit distribution scheme to obtain a simulation data package for each calculation unit; starting parallel electromagnetic transient calculation based on the simulation data package of each calculation unit and establishing a boundary data exchange communication mechanism to obtain electromagnetic transient simulation results.

2. The data partitioning method of an electromagnetic simulation model of a power system according to claim 1, wherein, The S2 step comprises: determining the cluster number K value based on the electrical distance matrix, and setting the cluster parameters of the K-means clustering algorithm; inputting the electrical distance values in the electrical distance matrix as a similarity measurement standard into the K-means clustering algorithm for initial cluster center selection to obtain initial cluster center coordinates; performing distance calculation and classification allocation processing on each node of the power system according to the initial cluster center coordinates and cluster parameters to obtain a node clustering grouping result; updating the cluster center position of the node clustering grouping result by iteration optimization to obtain the plurality of electrically coupled subsystems. 3.The data splitting processing method of an electromagnetic simulation model of a power system according to claim 1, characterized in that, The S3 step comprises: performing topological connection analysis based on the node attribution relationship of the plurality of electrically coupled subsystems to obtain a connection topological structure between subsystems; identifying the connection nodes across different electrically coupled subsystems according to the connection topological structure between subsystems to obtain a boundary node list; performing branch connection relationship retrieval on the nodes in the boundary node list to obtain boundary branches connecting different electrically coupled subsystems; The boundary node list and the boundary branch connecting different electrical coupling subsystems are matched and combined to obtain boundary nodes and boundary branches of each electrical coupling subsystem. 4.The data partitioning method of an electromagnetic simulation model of a power system according to claim 1, wherein, The S4 step comprises: Based on the boundary nodes and boundary branches of each electrical coupling subsystem, the Thevenin equivalent voltage source and equivalent impedance parameters are calculated to obtain boundary equivalent parameters. According to the boundary equivalent parameters, the boundary connection of each electrical coupling subsystem is replaced to obtain an independent subsystem model. The node admittance matrix is grouped and split according to the nodes of each electrical coupling subsystem to obtain the corresponding admittance sub-matrix of each electrical coupling subsystem. Based on the boundary equivalent parameters and the admittance sub-matrix, data combination and packaging are performed to obtain independent admittance sub-matrix and equivalent boundary parameters of each electrical coupling subsystem.

5. The data partitioning method of an electromagnetic simulation model of a power system according to claim 4, wherein, The node admittance matrix is grouped and split according to the nodes of each electrical coupling subsystem to obtain the corresponding admittance sub-matrix of each electrical coupling subsystem, comprising: Based on the node belonging relationship of each electrical coupling subsystem, the row and column indexes of the node admittance matrix are sorted to obtain a reordered node admittance matrix. According to the number of nodes of each electrical coupling subsystem, the matrix block boundary position is determined to obtain a matrix block index coordinate. The reordered node admittance matrix is extracted according to the matrix block index coordinate to obtain the original admittance sub-matrix of each electrical coupling subsystem. The original admittance sub-matrix of each electrical coupling subsystem is subjected to matrix integrity inspection and data format standardization processing to obtain the corresponding admittance sub-matrix of each electrical coupling subsystem.

6. The data partitioning method of an electromagnetic simulation model of a power system according to claim 5, wherein, According to the number of nodes of each electrical coupling subsystem, the matrix block boundary position is determined to obtain a matrix block index coordinate, comprising: The number of nodes contained in each electrical coupling subsystem is counted to obtain the node count value of each electrical coupling subsystem. Based on the node count value of each electrical coupling subsystem, an accumulated calculation is performed to obtain the starting position index of each electrical coupling subsystem in the matrix. According to the starting position index and node count value of each electrical coupling subsystem in the matrix, the termination position index of each subsystem is calculated to obtain the matrix boundary range of each electrical coupling subsystem. The starting position index and matrix boundary range of each electrical coupling subsystem are combined to obtain the matrix block index coordinate.

Images