CN-121997566-A - Hierarchical data driving modeling method and modeling device for large-scale power system

CN121997566ACN 121997566 ACN121997566 ACN 121997566ACN-121997566-A

Abstract

The application relates to the technical field of modeling and simulation of power systems, in particular to a hierarchical data driving modeling method and modeling device for a large-scale power system, wherein the method comprises the following steps: decoupling a target large-scale power system into a dynamic element set and an internet, then establishing an element model to determine a corresponding state space matrix, then establishing a network model to generate a network matrix, thereby aggregating the element model state space matrix to generate a block diagonal aggregation matrix, and constructing a global system state matrix by utilizing a preset analytic fusion formula. Therefore, the problems that in the related technology, the model complexity and the training data quantity exponentially increase along with the quantity of system state variables, the subsystem is divided depending on accurate physical topology and manual operation, a large number of element model parameters are required to be collected and integrated, and an integrated frame with complete systematic theory is lacking, so that the modeling cost is high, the generalization and the expandability are poor, and the overall dynamic characteristics of the system are difficult to accurately reflect are solved.

Inventors

LIU WEICHENG
LU CHAO
LIU DI

Assignees

清华大学

Dates

Publication Date: 20260508
Application Date: 20251230

Claims (10)

1. The hierarchical data driving modeling method for the large-scale power system is characterized by comprising the following steps of: decoupling a target large-scale power system into a dynamic element set and an internet; establishing a data-driven element model corresponding to each dynamic element of the dynamic element set to determine a corresponding state space matrix based on the element model and local operation conditions of the dynamic elements; establishing a network model of the Internet to describe algebraic relations among port variables of the dynamic element set, and generating a network matrix; And aggregating the element model state space matrix of each dynamic element to generate a system-level block diagonal aggregation matrix, and combining the block diagonal aggregation matrix and the network matrix by utilizing a preset analytic fusion formula to construct a single global system state matrix describing the overall dynamic state of the target large-scale power system.
2. The method of claim 1, wherein the predetermined analytical fusion formula is: , Wherein, the For the global system state matrix, To describe the block diagonal aggregation matrix of dynamic element internal evolution characteristics, To describe the block diagonal aggregation matrix of the interface inputs that affects the characteristics of the internal states, For the matrix of the network to be described, To describe the block diagonal aggregation matrix of interface input versus interface output direct transfer characteristics, The block diagonal aggregation matrix is output mapping characteristics for interfaces describing internal states.
3. The method of claim 1, wherein the building a data-driven element model for each dynamic element of the set of dynamic elements comprises: And performing collaborative training on all the element models by using a preset multi-objective loss function.
4. A method according to claim 3, wherein the pre-set multi-objective loss function comprises a local loss term for characterizing the prediction accuracy of the model and a global loss term for characterizing the accuracy of the overall dynamic response of the system predicted by the global system state matrix.
5. The method of claim 4, wherein the pre-set multi-objective loss function further comprises a network loss term that characterizes a prediction accuracy of the network model.
6. The method of claim 1, wherein the generating a network matrix comprises: Acquiring network topology and line impedance parameters; And obtaining a node admittance matrix based on the network topology, the line impedance parameters and the algebraic relation, and taking the node admittance matrix as the network matrix.
7. A hierarchical data-driven modeling apparatus for a large-scale power system, comprising: The decoupling module is used for decoupling the target large-scale power system into a dynamic element set and an internet; The first establishing module is used for establishing a data-driven element model corresponding to each dynamic element of the dynamic element set so as to determine a corresponding state space matrix based on the element model and the local operation condition of the dynamic element; the second building module is used for building a network model of the internet so as to describe algebraic relations among port variables of the dynamic element set and generate a network matrix; And the third building module is used for aggregating the element model state space matrix of each dynamic element to generate a system-level block diagonal aggregation matrix, and combining the block diagonal aggregation matrix and the network matrix by utilizing a preset analytic fusion formula to construct a single global system state matrix for describing the overall dynamic state of the target large-scale power system.
8. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor executing the program to implement the hierarchical data-driven modeling method for a large-scale power system of any of claims 1-6.
9. A computer readable storage medium having stored thereon a computer program, characterized in that the program is executed by a processor for implementing the hierarchical data driven modeling method for a large-scale power system according to any of claims 1-6.
10. A computer program product comprising a computer program, characterized in that the computer program is executed for implementing a hierarchical data driven modeling method for a large-scale power system according to any of claims 1-6.

Description

Hierarchical data driving modeling method and modeling device for large-scale power system Technical Field The application relates to the technical field of modeling and simulation of power systems, in particular to a hierarchical data driving modeling method and modeling device for a large-scale power system. Background Accurate and efficient dynamic modeling is a precondition and core foundation for supporting safe and stable operation of a power grid. The number of novel dynamic elements such as wind power, photovoltaic and other renewable energy sources is increased sharply, the system scale is expanded continuously, novel characteristics such as bidirectional power flow, multi-time scale coupling, complex control logic and the like are brought, the nonlinearity and uncertainty of the dynamic behavior of the power grid are enhanced remarkably, and a plurality of students are led to study on a modeling method of a large-scale power system. In the related art, a modeling method for a large-scale power system is a centralized data driving method, wherein the whole large-scale power system is regarded as a single huge dynamic system, a global dynamic state of the system is directly learned from end to end by means of a large-scale neural network model such as huge NODE (Neural Ordinary Differential Equations, nerve differential equation), a decomposition coordination algorithm is used for modeling by means of precisely grasping and manually dividing subsystems of a physical topology of the system and then completing modeling by means of a coordination subsystem model, a mechanism modeling method is used for constructing a system model by collecting and integrating precise models and parameters of hundreds of elements in the system, and a data driving mode is adopted for a single dynamic element by means of an element level modeling method. However, in the related art, because the centralized data driving method adopts a 'large-integration' model architecture, the complexity of the model, the required training data amount and the training calculation cost are exponentially increased along with the number of system state variables, serious 'dimension disasters' are encountered, the model is easy to overfit and poor in generalization performance, because the decomposition coordination algorithm depends on accurate physical topology and manual division subsystems, the division mode is stiff, flexibility is lacking, complicated iterative calculation is required during coordination, convergence and calculation efficiency are difficult to ensure, because the mechanism modeling method needs to collect accurate models and parameters for integrating a large number of elements, modeling work is tedious and time-consuming, local element parameter errors or loss can cause huge deviation of the whole system model, model maintenance and updating are extremely difficult, because the element-level modeling method stops at an element level, and lacks a systematic and theoretical complete integration frame, the scattered element model which is independently learned cannot be assembled into a coordinated global model reflecting the overall dynamic characteristics of the system, and the spanning from 'element intelligence' to 'system intelligence' is difficult to realize. Disclosure of Invention The application provides a hierarchical data driving modeling method and modeling device for a large-scale power system, which are used for solving the problems that modeling cost is high, generalization and expandability are poor, and overall dynamic characteristics of the system are difficult to accurately reflect due to the fact that model complexity and training data quantity in related technologies exponentially increase along with the number of system state variables, a large number of element model parameters are required to be collected and integrated depending on accurate physical topology and manual dividing subsystems, and a complete integration frame of systematic theory is lacked. An embodiment of the first aspect of the application provides a hierarchical data driving modeling method for a large-scale power system, which comprises the steps of decoupling a target large-scale power system into a dynamic element set and an internet, establishing a data driving element model corresponding to each dynamic element of the dynamic element set to determine a corresponding state space matrix based on the element model and local operation conditions of the dynamic elements, establishing a network model of the internet to describe algebraic relations among port variables of the dynamic element set to generate a network matrix, aggregating the element model state space matrix of each dynamic element to generate a system-level block diagonal aggregation matrix, and combining the block diagonal aggregation matrix and the network matrix by utilizing a preset analytic fusion formula to construct a single global system state matrix describ