Hierarchical data driven modeling method and modeling device for large-scale power system

Abstract

The invention relates to the technical field of power system modeling and simulation, in particular to a hierarchical data-driven modeling method and modeling device for a large-scale power system.The method comprises the steps that a target large-scale power system is decoupled into a dynamic element set and an internet, then an element model is established, a corresponding state space matrix is determined, and the dynamic element set and the internet are established; and then establishing a network model to generate a network matrix, aggregating the element model state space matrix to generate a block diagonal aggregation matrix, and constructing a global system state matrix by using a preset analysis fusion formula. Therefore, the problems of high modeling cost, poor generalization and expandability, high efficiency and the like caused by exponential increase of model complexity and training data volume along with the number of system state variables, dependence on precise physical topology and manual subsystem division, need of collecting and integrating a large number of element model parameters and lack of an integration framework with complete systematic theories in related technologies are solved. And the overall dynamic characteristics of the system are difficult to reflect accurately.

Description

Technical Field

[0001] This application relates to the field of power system modeling and simulation technology, and in particular to a hierarchical data-driven modeling method and modeling device for large-scale power systems. Background Technology

[0002] Accurate and efficient dynamic modeling is the prerequisite and core foundation for supporting the safe and stable operation of the power grid. The large-scale grid connection of renewable energy sources such as wind power and photovoltaics, and the rapid increase in the number of new dynamic components such as power electronic energy storage, have not only led to a continuous expansion of the system scale, but also brought new characteristics such as bidirectional power flow, multi-timescale coupling, and complex control logic. This has resulted in a significant increase in the nonlinearity and uncertainty of the power grid's dynamic behavior, prompting many scholars to study modeling methods for large-scale power systems.

[0003] Among related technologies, modeling methods for large-scale power systems fall into four categories: First, centralized data-driven methods, which attempt to treat the entire large-scale power system as a single, massive dynamic system and use a unified neural network model, such as a giant NODE (Neural Ordinary Differential Equations), to directly learn the global dynamics of the system end-to-end. Second, decomposition and coordination algorithms, which rely on a precise understanding of the system's physical topology and manual division into subsystems, and then complete the modeling by coordinating the subsystem models. Third, mechanistic modeling methods, which construct a system model by collecting and integrating precise models and parameters of hundreds or thousands of components within the system. Fourth, component-level modeling methods, which use a data-driven approach to model individual dynamic components.

[0004] However, in related technologies, centralized data-driven methods employ a "unified" model architecture, leading to an exponential increase in model complexity, required training data volume, and training computation costs with the number of system state variables, resulting in a severe "curse of dimensionality." Furthermore, these models are prone to overfitting and exhibit poor generalization performance. Decomposition and coordination algorithms rely on precise physical topology and manual subsystem partitioning, resulting in rigid partitioning methods lacking flexibility. Moreover, complex iterative calculations during coordination make it difficult to guarantee convergence and computational efficiency. Mechanistic modeling methods require the collection and integration of precise models and parameters for a large number of components, leading to tedious and time-consuming modeling work. Errors or missing parameters in local components can cause significant deviations in the entire system model, making model maintenance and updates extremely difficult. Finally, component-level modeling methods stop at the component level, lacking a systematic and theoretically complete integration framework. This prevents the "assembly" of independently learned, fragmented component models into a coordinated global model reflecting the overall dynamic characteristics of the system, hindering the leap from "component intelligence" to "system intelligence," which urgently needs to be addressed. Summary of the Invention

[0005] This application provides a hierarchical data-driven modeling method and modeling device for large-scale power systems to solve the problems in related technologies, such as the exponential growth of model complexity and training data volume with the number of system state variables, reliance on accurate physical topology and manual subsystem division, the need to collect and integrate a large number of component model parameters, and the lack of a systematic and complete integration framework, which leads to high modeling costs, poor generalization and scalability, and difficulty in accurately reflecting the overall dynamic characteristics of the system.

[0006] The first aspect of this application provides a hierarchical data-driven modeling method for large-scale power systems, comprising the following steps: decoupling the target large-scale power system into a set of dynamic components and an interconnection network; establishing a data-driven component model corresponding to each dynamic component in the set of dynamic components, and determining the corresponding state space matrix based on the component model and the local operating conditions of the dynamic component; establishing a network model of the interconnection network to describe the algebraic relationship between the port variables of the set of dynamic components, and generating a network matrix; aggregating the state space matrix of the component model of each dynamic component to generate a system-level block diagonal aggregation matrix, and combining the block diagonal aggregation matrix and the network matrix using a preset analytical fusion formula to construct a single global system state matrix describing the overall dynamics of the target large-scale power system.

[0007] Through the above technical means, the embodiments of this application can decouple the system and build component models in parallel, so that the modeling complexity, training data volume and system scale increase linearly, reducing the modeling cost. At the same time, it does not need to rely on the precise physical topology of the entire system or manually divide subsystems. It only needs to build a network model through port variable relationships, which greatly reduces the difficulty of the early preparation of modeling and adapts to system scenarios with different topologies. In addition, a theoretically complete integrated framework can be built through the hierarchical process of component modeling, network modeling and aggregation fusion. The global system state matrix is ​​generated by analysis and fusion, which accurately reflects the overall dynamic characteristics of the system. In addition, when adding or replacing dynamic components, only the corresponding component model needs to be updated, which improves maintenance efficiency. Moreover, the global system state matrix has a clear physical meaning and can accurately reflect the overall dynamic characteristics of the system. It can be directly connected to classical control theory and effectively improve the generalization performance of the model.

[0008] Optionally, in one embodiment of this application, the preset analytical fusion formula may be, but is not limited to, the following: , in, For the global system state matrix, A block diagonal aggregation matrix describing the internal evolution characteristics of dynamic elements. This is a block diagonal aggregation matrix describing the characteristics of how interface inputs affect the internal state. For network matrix, This is a block diagonal aggregation matrix describing the characteristics of direct transfer of interface input to interface output. This is a block diagonal aggregation matrix that describes the mapping characteristics of internal states to interface outputs.

[0009] Through the above technical means, the embodiments of this application can combine the block diagonal aggregation matrix and the network matrix by pre-set analytical fusion formula, using only mathematical operations and without iterative processes, effectively improving the integration efficiency of large-scale power system dynamic models. At the same time, by eliminating port variables and retaining only the internal state variables of the system, the global dynamic equations can accurately reflect the core dynamic characteristics of the system, reduce the errors introduced by intermediate variables, and improve the modeling accuracy.

[0010] Optionally, in one embodiment of this application, establishing a data-driven element model corresponding to each dynamic element in the dynamic element set includes: performing collaborative training on all element models using a preset multi-objective loss function.

[0011] Through the above technical means, the embodiments of this application can perform collaborative training on all component models, so that the component models have both individual dynamic fitting accuracy and cross-component interaction compatibility, avoiding conflicts when aggregating the state space matrix of each dynamic component model in the future.

[0012] Optionally, in one embodiment of this application, the preset multi-objective loss function includes a local loss term and a global loss term. The local loss term is used to characterize the prediction accuracy of the model, and the global loss term is used to characterize the accuracy of the overall dynamic response of the system predicted by the global system state matrix.

[0013] Through the above technical means, the embodiments of this application can use local loss terms to ensure the individual prediction accuracy of each dynamic element model, providing reliable element-level parameter support for global modeling. At the same time, by using global loss terms, the dynamic accuracy of the global system is incorporated into the optimization objective, so that the overall dynamic response of the system is highly consistent with the actual operating conditions, thereby improving the efficiency and reliability of dynamic modeling of large-scale power systems.

[0014] Optionally, in one embodiment of this application, the preset multi-objective loss function further includes a network loss term, which is used to characterize and measure the prediction accuracy of the network model.

[0015] Through the above technical means, the embodiments of this application can directly ensure the prediction accuracy of the network model by utilizing the network loss term, avoiding the ineffective connection between the component model and the network layer due to network model deviation. At the same time, through the linkage constraint of the network loss term, local loss term and global loss term, the accuracy of the component model and the network model forms a positive synergy, reducing the conflict when fusing the block diagonal aggregation matrix and the network matrix in the later stage, and enhancing the reliability of the model.

[0016] Optionally, in one embodiment of this application, generating the network matrix includes: obtaining network topology and line impedance parameters; and obtaining a node admittance matrix based on the network topology, the line impedance parameters, and the algebraic relationship, as the network matrix.

[0017] Through the above technical means, the embodiments of this application can generate a node admittance matrix based on network topology and line impedance parameters, avoiding the problems of scattered parameter collection and cumbersome integration in related mechanism modeling, ensuring that the network matrix can accurately reflect the real electrical characteristics of the interconnected network, and reducing modeling errors caused by parameter deviations.

[0018] A second aspect of this application provides a hierarchical data-driven modeling apparatus for large-scale power systems, comprising: a decoupling module for decoupling a target large-scale power system into a set of dynamic components and an interconnection network; a first establishment module for establishing a data-driven component model corresponding to each dynamic component of the set of dynamic components, and determining the corresponding state space matrix based on the component model and the local operating conditions of the dynamic component; a second establishment module for establishing a network model of the interconnection network to describe the algebraic relationship between the port variables of the set of dynamic components and generate a network matrix; and a third establishment module for aggregating the state space matrix of the component model of each dynamic component to generate a system-level block diagonal aggregation matrix, and combining the block diagonal aggregation matrix and the network matrix using a preset analytical fusion formula to construct a single global system state matrix describing the overall dynamics of the target large-scale power system.

[0019] Through the above technical means, the embodiments of this application can decouple the system and build component models in parallel, so that the modeling complexity, training data volume and system scale increase linearly, reducing the modeling cost. At the same time, it does not need to rely on the precise physical topology of the entire system or manually divide subsystems. It only needs to build a network model through port variable relationships, which greatly reduces the difficulty of the early preparation of modeling and adapts to system scenarios with different topologies. In addition, a theoretically complete integrated framework can be built through the hierarchical process of component modeling, network modeling and aggregation fusion. The global system state matrix is ​​generated by analysis and fusion, which accurately reflects the overall dynamic characteristics of the system. In addition, when adding or replacing dynamic components, only the corresponding component model needs to be updated, which improves maintenance efficiency. Moreover, the global system state matrix has a clear physical meaning and can accurately reflect the overall dynamic characteristics of the system. It can be directly connected to classical control theory and effectively improve the generalization performance of the model.

[0020] Optionally, in one embodiment of this application, the preset analytical fusion formula may be, but is not limited to, the following: , in, For the global system state matrix, A block diagonal aggregation matrix describing the internal evolution characteristics of dynamic elements. This is a block diagonal aggregation matrix describing the characteristics of how interface inputs affect the internal state. For network matrix, This is a block diagonal aggregation matrix describing the characteristics of direct transfer of interface input to interface output. This is a block diagonal aggregation matrix that describes the mapping characteristics of internal states to interface outputs.

[0021] Through the above technical means, the embodiments of this application can combine the block diagonal aggregation matrix and the network matrix by pre-set analytical fusion formula, using only mathematical operations and without iterative processes, effectively improving the integration efficiency of large-scale power system dynamic models. At the same time, by eliminating port variables and retaining only the internal state variables of the system, the global dynamic equations can accurately reflect the core dynamic characteristics of the system, reduce the errors introduced by intermediate variables, and improve the modeling accuracy.

[0022] Optionally, in one embodiment of this application, the first establishment module includes: a training unit, used to collaboratively train all component models using a preset multi-objective loss function.

[0023] Through the above technical means, the embodiments of this application can perform collaborative training on all component models, so that the component models have both individual dynamic fitting accuracy and cross-component interaction compatibility, avoiding conflicts when aggregating the state space matrix of each dynamic component model in the future.

[0024] Optionally, in one embodiment of this application, the preset multi-objective loss function includes a local loss term and a global loss term. The local loss term is used to characterize the prediction accuracy of the model, and the global loss term is used to characterize the accuracy of the overall dynamic response of the system predicted by the global system state matrix.

[0025] Through the above technical means, the embodiments of this application can use local loss terms to ensure the individual prediction accuracy of each dynamic element model, providing reliable element-level parameter support for global modeling. At the same time, by using global loss terms, the dynamic accuracy of the global system is incorporated into the optimization objective, so that the overall dynamic response of the system is highly consistent with the actual operating conditions, thereby improving the efficiency and reliability of dynamic modeling of large-scale power systems.

[0026] Optionally, in one embodiment of this application, the preset multi-objective loss function further includes a network loss term, which is used to characterize and measure the prediction accuracy of the network model.

[0027] Through the above technical means, the embodiments of this application can directly ensure the prediction accuracy of the network model by utilizing the network loss term, avoiding the ineffective connection between the component model and the network layer due to network model deviation. At the same time, through the linkage constraint of the network loss term, local loss term and global loss term, the accuracy of the component model and the network model forms a positive synergy, reducing the conflict when fusing the block diagonal aggregation matrix and the network matrix in the later stage, and enhancing the reliability of the model.

[0028] Optionally, in one embodiment of this application, the second establishing module includes: an acquisition unit for acquiring network topology and line impedance parameters; and a generation unit for obtaining a node admittance matrix based on the network topology, the line impedance parameters, and the algebraic relationship, as the network matrix.

[0029] Through the above technical means, the embodiments of this application can generate a node admittance matrix based on network topology and line impedance parameters, avoiding the problems of scattered parameter collection and cumbersome integration in related mechanism modeling, ensuring that the network matrix can accurately reflect the real electrical characteristics of the interconnected network, and reducing modeling errors caused by parameter deviations.

[0030] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the hierarchical data-driven modeling method for large-scale power systems as described in the above embodiments.

[0031] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the hierarchical data-driven modeling method for large-scale power systems as described above.

[0032] A fifth aspect of this application provides a computer program product, including a computer program that, when executed, implements the hierarchical data-driven modeling method for large-scale power systems as described above.

[0033] This application's embodiments achieve system decoupling and parallel construction of component models, allowing modeling complexity, training data volume, and system scale to grow linearly, reducing modeling costs. Simultaneously, it eliminates the need for precise physical topology or manual subsystem partitioning, constructing network models solely through port variable relationships, significantly reducing pre-modeling preparation difficulty and adapting to system scenarios with different topologies. Furthermore, a theoretically complete integration framework is built through a layered process of component modeling, network modeling, and aggregation fusion. A global system state matrix is ​​generated through analytical fusion, accurately reflecting the overall dynamic characteristics of the system. Additionally, when adding or replacing dynamic components, only the corresponding component model needs updating, improving maintenance efficiency. Moreover, the global system state matrix has clear physical meaning, accurately reflecting the overall dynamic characteristics of the system, and can directly interface with classical control theory, effectively improving model generalization performance. Therefore, this solves the problems in related technologies where model complexity and training data volume increase exponentially with the number of system state variables, relying on precise physical topology and manual subsystem partitioning, requiring the collection and integration of a large number of component model parameters, lacking a systematically theoretically complete integration framework, resulting in high modeling costs, poor generalization and scalability, and difficulty in accurately reflecting the overall dynamic characteristics of the system.

[0034] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0035] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a schematic diagram of the overall architecture of layered modeling provided according to an embodiment of this application; Figure 2 This is a flowchart illustrating a hierarchical data-driven modeling method for large-scale power systems according to an embodiment of this application. Figure 3 This is a flowchart illustrating the principle of the hierarchical data-driven modeling method for large-scale power systems provided in the embodiments of this application; Figure 4 This is a block diagram of a hierarchical data-driven modeling apparatus for large-scale power systems provided according to an embodiment of this application; Figure 5 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of this application. Detailed Implementation

[0036] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0037] The following description, with reference to the accompanying drawings, illustrates a hierarchical data-driven modeling method and apparatus for large-scale power systems according to embodiments of this application. Addressing the issues mentioned in the background art, such as the exponential growth of model complexity and training data volume with the number of system state variables, reliance on precise physical topology and manual subsystem partitioning, the need to collect and integrate a large number of component model parameters, and the lack of a systematic and theoretically complete integration framework, which leads to high modeling costs, poor generalization and scalability, and difficulty in accurately reflecting the overall dynamic characteristics of the system, this application provides a hierarchical data-driven modeling method for large-scale power systems. In this method, system decoupling and parallel construction of component models allow for a linear increase in modeling complexity and training data volume with system size, reducing modeling costs. Simultaneously, it eliminates the need to rely on system-wide modeling. By integrating precise physical topology or manually dividing subsystems, and constructing network models solely through port variable relationships, the difficulty of pre-modeling preparation is significantly reduced. This approach adapts to system scenarios with different topologies. Furthermore, a hierarchical process of component modeling, network modeling, and aggregation fusion can build a theoretically complete integrated framework. A global system state matrix is ​​generated through analytical fusion, accurately reflecting the overall dynamic characteristics of the system. Additionally, when adding or replacing dynamic components, only the corresponding component model needs to be updated, improving maintenance efficiency. The global system state matrix has clear physical meaning, accurately reflecting the overall dynamic characteristics of the system, and can directly interface with classical control theory, effectively improving model generalization performance. This solves the problems in related technologies where model complexity and training data volume increase exponentially with the number of system state variables, relying on precise physical topology and manually dividing subsystems, requiring the collection and integration of a large number of component model parameters, lacking a systematic and theoretically complete integrated framework, resulting in high modeling costs, poor generalization and scalability, and difficulty in accurately reflecting the overall dynamic characteristics of the system.

[0038] Before introducing the hierarchical data-driven modeling method for large-scale power systems provided in the embodiments of this application, we will first give a brief introduction to the overall architecture of hierarchical modeling provided in the embodiments of this application.

[0039] It is understood that the embodiments of this application aim to decompose a high-dimensional, complex global system modeling task into multiple low-dimensional, decoupled local modeling tasks, and finally integrate the local models into a global system model with clear physical meaning and rigorous mathematical form through an analytical method.

[0040] Figure 1This is a schematic diagram of the overall architecture of layered modeling provided according to an embodiment of this application.

[0041] Step S101: Parallel modeling of the component layer.

[0042] In this embodiment, parallel modeling at the component level is possible. Specifically, since the global operating condition is the operating scenario of the dynamic component, it can be used as the basic input for modeling. Then, through the operating condition mapping network N_map,1,...,N_map,N, the global operating condition is transformed into the local operating conditions p_d,1,...,p_d,N for each dynamic component. Subsequently, component models 1,...,N are generated based on the local operating conditions p_d,1,...,p_d,N. Furthermore, the component models can calculate the port output based on the port input and internal state of a single dynamic component.

[0043] Step S102: Network layer modeling.

[0044] In this embodiment, network layer modeling can be performed to generate a network model M_net. Furthermore, the network model can impose constraints on the port inputs and outputs of all dynamic components.

[0045] Step S103: Global model analysis and fusion.

[0046] In this embodiment, the global model parsing and fusion of the component model and the network model can be performed in the parsing fusion unit to generate the global system state matrix A_sys.

[0047] Specifically, Figure 2 This is a flowchart of a hierarchical data-driven modeling method for large-scale power systems provided according to an embodiment of this application.

[0048] like Figure 2 As shown, this hierarchical data-driven modeling method for large-scale power systems includes the following steps: In step S201, the target large-scale power system is decoupled into a dynamic set of components and an interconnected network.

[0049] In the embodiments of this application, the target large-scale power system refers to a large-scale power system that requires accurate and efficient dynamic modeling. The large scale can be reflected in the large number of dynamic components, the large number of system state variables, and the complex network topology.

[0050] In actual implementation, in order to simplify the complexity of the modeling task, the embodiments of this application can structurally decouple the target large-scale power system into two layers: the component layer and the network layer.

[0051] The component layer can be understood as a collection of dynamic components, including all independent dynamic components in the system, such as generators, inverters, and dynamic loads.

[0052] The network layer can be understood as an interconnected network, describing the topology and electrical characteristics of the power network that connects all dynamic components. Essentially, it is a set of algebraic constraint equations.

[0053] For example, embodiments of this application may be based on Figure 1 The overall architecture of the hierarchical modeling shown decomposes the target large-scale power system into a dynamic set of components (Device 1, ..., Device N) and an interconnection network (Power Network).

[0054] In step S202, a data-driven component model is established for each dynamic component in the dynamic component set, so as to determine the corresponding state space matrix based on the component model and the local operating conditions of the dynamic component.

[0055] In the embodiments of this application, the component model refers to a mathematical model that describes the dynamic response of a dynamic component, constructed based on the historical operating data of the dynamic component for each dynamic component in the dynamic component set.

[0056] In addition, local operating conditions refer to the operating state parameters of a dynamic element at a certain moment. They do not involve the global state of the entire power system and only reflect the current operating conditions of the dynamic element. For example, the local operating conditions of a generator include the current rotor speed, terminal voltage amplitude, and excitation current; the local operating conditions of an inverter include DC side voltage, AC current, and switching frequency; and the local operating conditions of a dynamic load include active power demand, reactive power demand, and voltage sensitivity.

[0057] Furthermore, the state space matrix can be understood as a parameter matrix that quantifies the state changes of dynamic elements. The state space matrix typically contains four fundamental sub-matrices: the state matrix... Input matrix Output matrix Direct transmission matrix The state matrix describes the evolution of the dynamic element's state when there is no external input, the input matrix describes the degree of influence of external input on the element's state, the output matrix describes the mapping relationship between the element's state and the output quantity, and the direct transfer matrix describes the direct influence of external input on the output quantity.

[0058] In actual implementation, the embodiments of this application can be applied to each dynamic element in the dynamic element set. ( Component models are built independently and in parallel using a data-driven approach. The goal of each component model is to learn a set of operating conditions from its local context. To its standard linear state space matrix The mapping relationship.

[0059] Based on the description of other embodiments, by way of example, this application embodiment uses an independent, data-driven modeling module to learn for each dynamic element Device i.

[0060] It should be noted that the modeling module can be an NSSE (Neural State Space Equation) model, which can learn a mapping based on the local operating conditions of the dynamic element. (Based on global operating conditions) Through a mapping network (Inferred), the small-signal dynamic characteristics of the dynamic element under this operating condition are generated using the state-space matrix. Represented in the form of .

[0061] Understandably, the above process is performed in parallel for all N dynamic elements, which greatly improves efficiency.

[0062] Optionally, in one embodiment of this application, establishing a data-driven element model corresponding to each dynamic element in the dynamic element set includes: performing collaborative training on all element models using a preset multi-objective loss function.

[0063] In actual implementation, the embodiments of this application can incorporate the model parameters of all dynamic elements into the same training framework through collaborative training. During the iterative optimization process, the preset multi-objective loss function is minimized simultaneously, so that each element model can actively adapt to the interaction constraints of other element models while learning its own dynamic laws. This avoids local optima and global conflicts caused by isolated training of a single dynamic element, and realizes dynamic collaboration between element models.

[0064] The embodiments of this application can perform collaborative training on all component models, so that the component models have both individual dynamic fitting accuracy and cross-component interaction compatibility, avoiding conflicts when aggregating the state space matrix of each dynamic component model in the later stage.

[0065] Optionally, in one embodiment of this application, the preset multi-objective loss function includes a local loss term and a global loss term. The local loss term is used to characterize the prediction accuracy of the model, and the global loss term is used to characterize the accuracy of the overall dynamic response of the system predicted by the global system state matrix.

[0066] In some cases, a pre-defined multi-objective loss function can achieve hierarchical optimization through the collaborative constraints of local and global loss terms.

[0067] Among them, the local loss term uses the actual operating data of a single dynamic element as a benchmark. It measures the individual prediction accuracy of the element model by calculating the deviation between the model prediction value and the actual value, so as to ensure that the local dynamic characteristics of each element model are accurately fitted.

[0068] The global loss term uses the actual overall operating data of a large-scale power system as a reference. It calculates the deviation between the overall dynamic response predicted by the global system state matrix and the actual global operating conditions, and constrains the accuracy of the contribution of the component model to the global dynamics.

[0069] It is understood that the embodiments of this application coordinately optimize the local loss term and the global loss term during the training process, so that while the component model ensures its own prediction accuracy, the global system state matrix formed by its aggregation can accurately reflect the overall dynamics of the system, avoiding the problem that the local model accuracy meets the standard but the global dynamic prediction fails.

[0070] The embodiments of this application can use local loss terms to ensure the individual prediction accuracy of each dynamic component model, providing reliable component-level parameter support for global modeling. At the same time, the global loss terms are used to incorporate the dynamic accuracy of the global system into the optimization objective, so that the overall dynamic response of the system is highly consistent with the actual operating conditions, thereby improving the efficiency and reliability of dynamic modeling of large-scale power systems.

[0071] Optionally, in one embodiment of this application, the preset multi-objective loss function further includes a network loss term, which is used to characterize and measure the prediction accuracy of the network model.

[0072] In some cases, a pre-defined multi-objective loss function can be used to quantify the prediction accuracy of the network model by calculating the deviation between the predicted and actual values ​​of these key electrical characteristics of the network model, based on the actual operating data of the interconnected network and using the network loss term.

[0073] Among them, the network loss term can work together with the local loss term and the global loss term and be incorporated into the optimization objective simultaneously. This allows the component model to adapt to the prediction rules of the network model during training, and the network model to match the dynamic output of the component model during optimization. This forms a linkage constraint between the component, the network, and the global model, avoiding the disconnect of dynamic characteristics caused by the separate optimization of the component model and the network model.

[0074] The embodiments of this application can directly guarantee the prediction accuracy of the network model by utilizing the network loss term, avoiding the ineffective connection between the component model and the network layer due to network model deviation. At the same time, through the linkage constraint of the network loss term, local loss term and global loss term, the accuracy of the component model and the network model forms a positive synergy, reducing the conflict when fusing the block diagonal aggregation matrix and the network matrix in the later stage, and enhancing the reliability of the model.

[0075] In step S203, a network model of the interconnected network is established to describe the algebraic relationship between the port variables of the dynamic element set, and a network matrix is ​​generated.

[0076] In the embodiments of this application, the network model refers to a mathematical model constructed based on the physical laws of power networks (such as Kirchhoff's laws) and containing algebraic constraint equations. These algebraic constraint equations can quantify the algebraic relationships between port variables of a dynamic element set and are used to describe the topology and electrical constraints of the interconnected network.

[0077] Furthermore, a network matrix refers to a quantified carrier that transforms the algebraic constraint equations in a network model into matrix-form mathematical parameters.

[0078] The embodiments of this application can establish a network model describing the algebraic constraints of an interconnected network, which defines the set of port output vectors of all components. With port input vector set The linear relationship between them can be represented as the network matrix. Therefore, the expression for the network matrix can be, but is not limited to, the following: , Among them, network matrix It can be constructed directly from known network topology and line parameters, or similarly learned from measurement data using a data-driven approach.

[0079] Specifically, in one embodiment of this application, generating a network matrix includes: obtaining network topology and line impedance parameters; and obtaining a node admittance matrix based on the network topology, line impedance parameters, and algebraic relationships, which serves as the network matrix.

[0080] In this embodiment, the network topology that characterizes the relationship between nodes and branches, and the line impedance parameters that reflect the electrical characteristics of the lines are first obtained. Then, based on electrical algebraic relationships, the network topology and line impedance parameters are transformed into a quantized node admittance matrix. Since the node admittance matrix can directly characterize the current and voltage relationship between nodes, it perfectly matches the network model's requirement for a quantized carrier of algebraic constraint equations. Therefore, the node admittance matrix is ​​used as the network matrix.

[0081] As can be understood from the description of other embodiments, the goal of the network model is to learn or define the network matrix. In some embodiments, when the network topology and line impedance parameters are known, These can be calculated in advance, such as the nodal admittance matrix; however, when the network topology or line impedance parameters are unknown, It can also be parameterized as a learnable matrix by minimizing the network loss. (Right now Identify it from the data.

[0082] The embodiments of this application can generate node admittance matrices based on network topology and line impedance parameters, avoiding the problems of scattered parameter collection and cumbersome integration in related mechanism modeling, ensuring that the network matrix can accurately reflect the real electrical characteristics of the interconnected network, and reducing modeling errors caused by parameter deviations.

[0083] In step S204, the element model state space matrix of each dynamic element is aggregated to generate a system-level block diagonal aggregate matrix. The block diagonal aggregate matrix and the network matrix are then combined using a preset analytical fusion formula to construct a single global system state matrix that describes the overall dynamics of the target large-scale power system.

[0084] In the embodiments of this application, the preset analytical fusion formula refers to the operation rule that merges the block diagonal aggregation matrix of the component layer and the network matrix of the network layer through pure algebraic operations (such as matrix addition and subtraction, inversion, left multiplication, right multiplication, etc.), which is used to eliminate component port variables and realize the integration of component dynamic characteristics and network topology constraints.

[0085] In addition, the global system state matrix refers to a quantitative matrix that describes the overall dynamic characteristics of a target large-scale power system. It is used to characterize the evolution of all internal state variables of the system and support applications such as system stability analysis and dynamic response prediction.

[0086] After obtaining the component model corresponding to the dynamic component and the network model corresponding to the interconnection network, this application embodiment can use analytical derivation to fuse all component models and network models to construct a global system state matrix.

[0087] Specifically, in this embodiment, the state space matrices of all N dynamic elements are first aggregated into four block diagonal aggregation matrices, the expressions of which may be, but are not limited to, as follows: , , , , in, , , , These correspond to A, B, C, and D, respectively: a block diagonal aggregation matrix describing the internal evolution characteristics of a dynamic element; a block diagonal aggregation matrix describing the influence of interface input on the internal state; a block diagonal aggregation matrix describing the mapping characteristics of the internal state to the interface output; and a block diagonal aggregation matrix describing the direct transmission characteristics of interface input to interface output. diag(·) is the constructor for the block diagonal aggregation matrix. For the first i The state matrix corresponding to each dynamic element For the first i The input matrix corresponding to each dynamic element No. i The output matrix corresponding to each dynamic element No. i The direct transmission matrix corresponding to each dynamic element i =1,..., N .

[0088] It should be noted that the core technology of the aforementioned operation of "aggregating the state space matrices of all N dynamic elements into four block diagonal aggregation matrices" lies in achieving the scalability of high-dimensional systems in large-scale scenarios through "physical partitioning".

[0089] Taking the power grid scenario as an example, the embodiments of this application can regard N dynamic components such as generators, energy storage devices or converters in the power grid as independent individuals, and combine the state space matrix of each dynamic component into a larger block matrix through the block diagonal operator.

[0090] This block matrix mathematically represents a "quasi-decoupled" state, meaning that, without considering network connections, the internal dynamic processes of each dynamic element, its sensitivity to interface signals, and its state mapping logic to the outside are all independent and parallel.

[0091] This approach is key to solving the "curse of dimensionality"—it doesn't directly rigidly fit a complex mapping at the global level, but instead first precisely characterizes the physical properties of each component at the local level. The most crucial correction lies in clarifying... It undertakes the function of information conversion "from the inside out," while It is responsible for the instantaneous feedback function of "from input to output".

[0092] The embodiments of this application, through the above-described modular block architecture, lay a structured foundation for the subsequent introduction of network topology constraints and their eventual fusion into a global system model.

[0093] Subsequently, embodiments of this application can use a preset analytical fusion formula to combine the block diagonal aggregation matrix and the network matrix. By performing analytical combination, the final global system state matrix describing the dynamics of the entire large-scale power system is obtained. Its expression can be, but is not limited to, as: .

[0094] Optionally, in one embodiment of this application, the preset analytical fusion formula may be, but is not limited to, the following: , in, For the global system state matrix, A block diagonal aggregation matrix describing the internal evolution characteristics of dynamic elements. This is a block diagonal aggregation matrix describing the characteristics of how interface inputs affect the internal state. For network matrix, This is a block diagonal aggregation matrix describing the characteristics of direct transfer of interface input to interface output. This is a block diagonal aggregation matrix that describes the mapping characteristics of internal states to interface outputs.

[0095] In actual implementation, the embodiments of this application can use a preset analytical fusion formula, which is mathematically equivalent to eliminating all port variables (such as voltage and current) of dynamic components and retaining only the internal state variables of all dynamic components. Thus, a global dynamic equation is established. .

[0096] The embodiments of this application can combine the block diagonal aggregation matrix and the network matrix through a preset analytical fusion formula. It only uses mathematical operations and does not require an iterative process, which effectively improves the integration efficiency of the dynamic model of a large-scale power system. At the same time, by eliminating port variables and retaining only the internal state variables of the system, the global dynamic equation can accurately reflect the core dynamic characteristics of the system, reduce the error introduced by intermediate variables, and improve the modeling accuracy.

[0097] Based on the descriptions of other embodiments, and exemplarily, embodiments of this application can perform end-to-end training of the entire hierarchical modeling framework, employing a preset multi-objective loss function during the training process. This function may include, but is not limited to, local loss terms. This is used to penalize the deviation between the predicted trajectory and the actual trajectory of each component model; the network loss term. This is used to penalize the deviation between the predicted output and the actual output of the network model; the global loss term. Used to penalize the global system state matrix obtained from the final fusion. The deviation between the predicted global state trajectory of the system and the actual trajectory.

[0098] In each iteration of training, embodiments of this application calculate the local loss term for each element model. This drives each NSSE model to accurately learn the dynamics of its corresponding components, and then, given that the network model is learnable, calculates the network loss term. This is to drive the network model to accurately match the algebraic constraints of the network, and then use the fused results. And ODESolve (Ordinary Differential Equation Solver), from an initial state Predict the global state trajectory of the entire system Thus, the global loss term is calculated. .

[0099] It should be noted that the total loss is the weighted sum of the local loss, network loss, and global loss. The expression can be, but is not limited to: , in, For network loss items The weighting coefficients, For local loss items The weighting coefficients, For global loss items The weighting coefficients.

[0100] It should be noted that the gradient comes from Backpropagation updates the learnable parameters in all component models simultaneously. Global loss. The existence of [the system] plays a crucial "coordination" role, forcing individual dynamic element models to be fine-tuned to ensure that the correct system-level emergent behavior can be reproduced when all dynamic elements are combined.

[0101] In summary, the embodiments of this application can transform a large-scale system identification problem into a manageable, scalable, and physically interpretable hierarchical learning and fusion task, ultimately yielding a high-precision global system model. It can be directly used for subsequent stability analysis. At the same time, by co-optimizing local loss terms, network loss terms and global loss terms, it can ensure that the component models are not only accurate on their own, but also compatible with each other and can be fused into an accurate global model.

[0102] The working principle of the hierarchical data-driven modeling method for large-scale power systems proposed in this application is illustrated below with a specific embodiment.

[0103] Figure 3 This is a flowchart illustrating the principle of a hierarchical data-driven modeling method for large-scale power systems provided in an embodiment of this application.

[0104] Step S301: Decouple the target large-scale power system into a dynamic set of components and an interconnected network.

[0105] Step S302: Establish the component model corresponding to each dynamic component in the dynamic component set, and determine the corresponding state space matrix based on the component model and the local operating conditions of the dynamic component.

[0106] Step S303: Aggregate the state space matrix corresponding to the dynamic element, and generate a block diagonal aggregation matrix based on the state space matrix.

[0107] Step S304: Establish a network model of the interconnected network and generate a network matrix.

[0108] Step S305: Combine the block diagonal aggregation matrix and the network matrix according to the preset analytical fusion formula.

[0109] In this application embodiment, a network matrix can be used as the basis. First, perform matrix subtraction on the block diagonal aggregation matrix D, which describes the direct transmission characteristics of the interface input to the interface output, to obtain... Perform the inverse operation to obtain Then, based on the block diagonal aggregation matrix C describing the mapping characteristics of the interface output to the internal state, matrix multiplication is performed to obtain... Then, based on the block diagonal aggregation matrix B describing the influence of the interface input on the internal state, matrix multiplication is performed to obtain... Then, based on the block diagonal aggregation matrix A describing the internal evolution characteristics of the dynamic element, matrix addition is performed to obtain... .

[0110] Step S306: Generate the global system state matrix.

[0111] The hierarchical data-driven modeling method for large-scale power systems proposed in this application can decouple the system and construct component models in parallel, making the modeling complexity and training data volume increase linearly with the system scale, thus reducing modeling costs. Simultaneously, it eliminates the need for precise physical topology of the entire system or manual subsystem partitioning, constructing a network model solely through port variable relationships, significantly reducing the difficulty of pre-modeling preparation and adapting to system scenarios with different topologies. Furthermore, a theoretically complete integration framework can be constructed through a hierarchical process of component modeling, network modeling, and aggregation and fusion. A global system state matrix is ​​generated through analytical fusion, accurately reflecting the overall dynamic characteristics of the system. Additionally, when adding or replacing dynamic components, only the corresponding component model needs to be updated, improving maintenance efficiency. Moreover, the global system state matrix has clear physical meaning, accurately reflecting the overall dynamic characteristics of the system, and can directly connect to classical control theory, effectively improving model generalization performance. Therefore, this solves the problems in related technologies where model complexity and training data volume increase exponentially with the number of system state variables, rely on precise physical topology and manual subsystem partitioning, require the collection and integration of a large number of component model parameters, lack a systematic and theoretically complete integration framework, resulting in high modeling costs, poor generalization and scalability, and difficulty in accurately reflecting the overall dynamic characteristics of the system.

[0112] Next, referring to the accompanying drawings, a hierarchical data-driven modeling apparatus for large-scale power systems is described according to an embodiment of this application.

[0113] Figure 4This is a block diagram of a hierarchical data-driven modeling apparatus for large-scale power systems provided according to an embodiment of this application.

[0114] like Figure 4 As shown, the hierarchical data-driven modeling device 40 for large-scale power systems includes: a decoupling module 100, a first establishment module 200, a second establishment module 300, and a third establishment module 400.

[0115] Among them, the decoupling module 100 is used to decouple the target large-scale power system into a dynamic set of components and an interconnected network.

[0116] The first establishment module 200 is used to establish a data-driven component model corresponding to each dynamic component in the dynamic component set, so as to determine the corresponding state space matrix based on the component model and the local operating conditions of the dynamic component.

[0117] The second module 300 is used to establish a network model of the interconnected network to describe the algebraic relationships between port variables of the dynamic component set and generate a network matrix.

[0118] The third module 400 is used to aggregate the element model state space matrix of each dynamic element to generate a system-level block diagonal aggregation matrix, and to combine the block diagonal aggregation matrix and the network matrix using a preset analytical fusion formula to construct a single global system state matrix that describes the overall dynamics of the target large-scale power system.

[0119] Optionally, in one embodiment of this application, the preset analytical fusion formula may be, but is not limited to, the following: , in, For the global system state matrix, A block diagonal aggregation matrix describing the internal evolution characteristics of dynamic elements. This is a block diagonal aggregation matrix describing the characteristics of how interface inputs affect the internal state. For network matrix, This is a block diagonal aggregation matrix describing the characteristics of direct transfer of interface input to interface output. This is a block diagonal aggregation matrix that describes the mapping characteristics of internal states to interface outputs.

[0120] Optionally, in one embodiment of this application, the first establishment module 200 includes: a training unit, used to collaboratively train all component models using a preset multi-objective loss function.

[0121] Optionally, in one embodiment of this application, the preset multi-objective loss function includes a local loss term and a global loss term. The local loss term is used to characterize the prediction accuracy of the model, and the global loss term is used to characterize the accuracy of the overall dynamic response of the system predicted by the global system state matrix.

[0122] Optionally, in one embodiment of this application, the preset multi-objective loss function further includes a network loss term, which is used to characterize and measure the prediction accuracy of the network model.

[0123] Optionally, in one embodiment of this application, the second establishment module 300 includes: an acquisition unit and a generation unit.

[0124] The acquisition unit is used to acquire network topology and line impedance parameters.

[0125] The generation unit is used to obtain the node admittance matrix based on the network topology, line impedance parameters, and algebraic relationships, which serves as the network matrix.

[0126] It should be noted that the foregoing explanation of the hierarchical data-driven modeling method embodiment for large-scale power systems also applies to the hierarchical data-driven modeling device for large-scale power systems in this embodiment, and will not be repeated here.

[0127] The hierarchical data-driven modeling device for large-scale power systems proposed in this application can decouple the system and construct component models in parallel, making the modeling complexity and training data volume increase linearly with the system scale, thus reducing modeling costs. Simultaneously, it eliminates the need for precise physical topology of the entire system or manual subsystem partitioning, constructing a network model solely through port variable relationships, significantly reducing the difficulty of pre-modeling preparation and adapting to system scenarios with different topologies. Furthermore, a theoretically complete integration framework can be constructed through a hierarchical process of component modeling, network modeling, and aggregation and fusion. A global system state matrix is ​​generated through analytical fusion, accurately reflecting the overall dynamic characteristics of the system. Additionally, when adding or replacing dynamic components, only the corresponding component model needs to be updated, improving maintenance efficiency. Moreover, the global system state matrix has clear physical meaning, accurately reflecting the overall dynamic characteristics of the system, and can directly connect to classical control theory, effectively improving model generalization performance. Therefore, it solves the problems in related technologies where model complexity and training data volume increase exponentially with the number of system state variables, relying on precise physical topology and manual subsystem partitioning, requiring the collection and integration of a large number of component model parameters, lacking a systematic and theoretically complete integration framework, resulting in high modeling costs, poor generalization and scalability, and difficulty in accurately reflecting the overall dynamic characteristics of the system.

[0128] Figure 5 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of this application. The electronic device may include: The memory 501, the processor 502, and the computer program stored on the memory 501 and capable of running on the processor 502.

[0129] When processor 502 executes the program, it implements the hierarchical data-driven modeling method for large-scale power systems provided in the above embodiments.

[0130] Furthermore, electronic devices also include: Communication interface 503 is used for communication between memory 501 and processor 502.

[0131] The memory 501 is used to store computer programs that can run on the processor 502.

[0132] Memory 501 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0133] If the memory 501, processor 502, and communication interface 503 are implemented independently, then the communication interface 503, memory 501, and processor 502 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0134] Optionally, in a specific implementation, if the memory 501, processor 502, and communication interface 503 are integrated on a single chip, then the memory 501, processor 502, and communication interface 503 can communicate with each other through an internal interface.

[0135] Processor 502 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0136] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the hierarchical data-driven modeling method for large-scale power systems as described above.

[0137] This application also provides a computer program product, including a computer program that, when executed, implements the hierarchical data-driven modeling method for large-scale power systems as described above.

[0138] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0139] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0140] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0141] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0142] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, it can be implemented using any one or more of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0143] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0144] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0145] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A hierarchical data-driven modeling method for large-scale power systems, characterized in that, Includes the following steps: Decouple the target large-scale power system into a dynamic collection of components and an interconnected network; Establish a data-driven component model for each dynamic component in the set of dynamic components, and determine the corresponding state space matrix based on the component model and the local operating conditions of the dynamic component. A network model of the interconnected network is established to describe the algebraic relationship between the port variables of the dynamic element set, and a network matrix is ​​generated. The element model state space matrix of each dynamic element is aggregated to generate a system-level block diagonal aggregate matrix. The block diagonal aggregate matrix and the network matrix are then combined using a preset analytical fusion formula to construct a single global system state matrix that describes the overall dynamics of the target large-scale power system.

2. The method according to claim 1, characterized in that, The preset analytical fusion formula is: ， in, The global system state matrix is... The block diagonal aggregation matrix describes the internal evolution characteristics of dynamic elements. The block diagonal aggregation matrix describes the characteristics of how interface inputs affect the internal state. The network matrix, The block diagonal aggregation matrix describes the direct transfer characteristics of interface input to interface output. The block diagonal aggregation matrix describes the mapping characteristics of internal states to interface outputs.

3. The method according to claim 1, characterized in that, The process of establishing a data-driven component model corresponding to each dynamic component in the dynamic component set includes: All component models are trained collaboratively using a pre-defined multi-objective loss function.

4. The method according to claim 3, characterized in that, The preset multi-objective loss function includes a local loss term and a global loss term. The local loss term is used to characterize the prediction accuracy of the model, and the global loss term is used to characterize the accuracy of the overall dynamic response of the system predicted by the global system state matrix.

5. The method according to claim 4, characterized in that, The preset multi-objective loss function also includes a network loss term, which is used to characterize and measure the prediction accuracy of the network model.

6. The method according to claim 1, characterized in that, The generated network matrix includes: Obtain network topology and line impedance parameters; The node admittance matrix is ​​obtained based on the network topology, the line impedance parameters, and the algebraic relationship, and serves as the network matrix.

7. A hierarchical data-driven modeling device for large-scale power systems, characterized in that, include: Decoupling modules are used to decouple a target large-scale power system into a dynamic collection of components and an interconnected network. The first establishment module is used to establish a data-driven element model corresponding to each dynamic element in the dynamic element set, so as to determine the corresponding state space matrix based on the element model and the local operating conditions of the dynamic element. The second establishment module is used to establish a network model of the interconnected network to describe the algebraic relationship between the port variables of the dynamic element set and generate a network matrix. The third module is used to aggregate the element model state space matrix of each dynamic element to generate a system-level block diagonal aggregation matrix, and to combine the block diagonal aggregation matrix and the network matrix using a preset analytical fusion formula to construct a single global system state matrix describing the overall dynamics of the target large-scale power system.

8. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the hierarchical data-driven modeling method for large-scale power systems as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the hierarchical data-driven modeling method for large-scale power systems as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, The computer program is executed to implement the hierarchical data-driven modeling method for large-scale power systems as described in any one of claims 1-6.

Images