A method, system and device for simulating power grid frequency propagation

Abstract

This invention provides a method, system, and apparatus for simulating power grid frequency propagation, relating to the field of power grid frequency propagation simulation technology. The method includes: Step 1, constructing a power grid frequency propagation model based on power grid dynamics theory; Step 2, constructing a PINN model; transforming the power grid dynamics equations into partial differential equation constraints and embedding them into the training process of the PINN model; Step 3, learning the operator mapping relationship between disturbance input and angular frequency response in the function space using the FNO operator; Step 4, jointly training the PINN model and the FNO operator to obtain a simulation model; Step 5, collecting the disturbance power value of the tested power grid and the random power output value of new energy sources, inputting them into the simulation model, predicting the angular frequency, and obtaining the frequency propagation curve. This scheme can accurately describe the power grid frequency dynamics under conditions of new energy power supply fluctuations, large disturbances, and complex boundary conditions.

Description

Technical Field

[0001] This invention relates to the field of power grid frequency propagation simulation technology, and in particular to a power grid frequency propagation simulation method, system and device. Background Technology

[0002] With the large-scale integration of new energy power supply and the increasing complexity of power systems, the dynamic stability of grid frequency and the problem of disturbance propagation are becoming increasingly prominent. Frequency deviation and oscillation not only threaten the safe operation of the power system, but may also lead to cascading reactions and large-scale power outages. Therefore, how to accurately and efficiently simulate the grid frequency propagation process has become an important research direction for power system analysis and dispatch control.

[0003] Existing power grid frequency propagation simulation methods mainly rely on traditional numerical modeling and simulation techniques. While these methods can describe the frequency dynamics of the system to some extent, they often face problems such as high computational overhead, insufficient real-time performance, and lack of physical consistency when dealing with new energy fluctuations, large disturbance conditions, and large-scale multi-node power grids. Traditional numerical simulations of difference algebraic equations (DAEs) rely on explicit / implicit integration algorithms. As the node size and simulation duration increase, the computational complexity rises exponentially. In power grids with hundreds of nodes or even larger scales, simulations require considerable time, making it difficult to meet the real-time requirements of scheduling, control, and emergency response plans. Lumped parameter models such as the swing equation, based on linearization assumptions, are effective under small disturbance conditions. However, in highly nonlinear and large disturbance scenarios such as frequent fluctuations in new energy sources, instability cutoff of high-power units, or short-circuit faults, the model error increases significantly and cannot accurately reflect the dynamic characteristics of the frequency. Multi-node coupling leads to strong spatiotemporal correlation in power grid frequency propagation. Traditional simplified modeling methods (such as mode order reduction and cluster equivalence) struggle to maintain prediction accuracy in large-scale, complex power grids, and require parameter readjustment under different operating conditions, lacking versatility. While deep learning methods (such as LSTM, GRU, and Transformer models) can predict frequency trends to some extent, their "black box" nature leads to: the model results lacking physical meaning and being difficult to interpret; poor generalization ability under unseen working conditions or disturbances; and susceptibility to noise and insufficient data, resulting in insufficient model stability. Existing methods perform reasonably well in traditional thermal power-dominated power grid environments, but as the proportion of new energy (wind power, photovoltaic) power supply increases, the propagation law of frequency disturbances becomes more complex due to reduced inertia and enhanced fluctuations, making it impossible for existing methods to balance accuracy and real-time performance. Moreover, the application of existing methods in actual power grid operation is mostly limited to offline analysis and planning, making it difficult to directly apply them to online frequency prediction, real-time regulation and emergency control. Summary of the Invention

[0004] The purpose of this invention is to provide a power grid frequency propagation simulation method, system, and apparatus to solve at least one of the aforementioned technical problems existing in the prior art.

[0005] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a power grid frequency propagation simulation method, comprising the following steps: Step 1: Construct a power grid frequency propagation model based on power grid dynamics theory.

[0006] In one feasible implementation, the power grid frequency propagation model includes power grid dynamics equations, power grid flow constraints, and disturbance terms; The power grid dynamics equation can be a generator swing equation, and the specific calculation formula includes: ; in, Indicates the first The inertial constant of each power grid node; Indicates the first Damping coefficient of each power grid node; Indicates the first The angular frequency of each power grid node; Indicates the rated angular frequency; Indicates the first Mechanical power input of each grid node; Indicates the first Electromagnetic output power of each power grid node; Indicates the total number of nodes in the power grid; Indicates time; The specific expressions for the power flow constraints include: ; in, Indicates the first Power flow values ​​of each grid node; Indicates the first Voltage amplitude at each grid node; Indicates the first Voltage amplitude at each grid node; It represents the conductance of branches between power grid nodes and is used to reflect active power loss; It represents the susceptance of branches between power grid nodes, used to reflect reactive power exchange and line charging effects; Indicates the first Voltage phase angle of each grid node With the Voltage phase angle of each grid node The difference between them, i.e. ; The disturbance term is used to realistically reflect the power grid operating conditions, so as to provide strict PDE constraints and physical priors for the training of subsequent neural networks, thereby ensuring the basic rationality of the model; the specific expression includes: ; in, Indicates the first Mechanical power input values ​​for each grid node; Indicates the first Initial values ​​of mechanical power of each grid node; This indicates the disturbance power value (such as disturbances caused by sudden power changes, generator loss of synchronization, line faults, etc.). This indicates the random output power value of the new energy source.

[0007] Step 2: Construct the PINN model (Physical Information Neural Network); transform the power grid dynamics equations into partial differential equation (PDE) constraints and embed them into the training process of the PINN model.

[0008] In one feasible implementation, the partial differential equation constraints include physical residual terms, boundary condition terms, and data error terms; The physical residual term represents the equation residual after substituting the angular frequency prediction data into the power grid dynamics equations. This ensures that the PINN model output satisfies the physical laws of the power system, thereby avoiding "black box" bias. The specific calculation formula includes: ; in, Represents the physical residual term; Indicates the number of sampling points for the physical residual term; The boundary condition terms are used to constrain the PINN model at the initial time (e.g., The solution at the rated frequency of 50 Hz (or at the boundary node) is used to maintain consistency with actual operating conditions. Specific expressions include: ; in, Represents boundary condition terms; Indicates the number of sampling points for the boundary condition term; This represents the initial value of the angular frequency prediction data; The data error term represents the difference between the predicted angular frequency data and the simulated (or measured) angular frequency data, in order to ensure that the PINN model can effectively fit the simulated (or measured) angular frequency data, thereby improving its adaptability to actual power grid conditions. The specific expression includes: ; in, Indicates the data error term; This represents the angular frequency prediction data; Indicates the first At that moment; Represents simulated (or measured) angular frequency data; This indicates the number of sampling points for the data error term; In this way, variables such as angular frequency, power, time, and grid node location can be used as inputs to the PINN model to predict the dynamic response of the grid node angular frequency. Three types of terms are introduced into the loss function: differential equation residuals, boundary condition residuals, and data errors. This ensures that the PINN model learning process conforms to physical laws and can fit real measurement or simulation data. Compared with pure data-driven methods, the PINN model significantly improves physical consistency and result interpretability.

[0009] Step 3: Using the FNO operator (Fourier neural operator), learn the operator mapping relationship between the perturbation input and the angular frequency response in the function space.

[0010] In one feasible implementation, the function space in step 3 refers to the set of continuous functions defined in the spatiotemporal domain, used to characterize the input-output relationship between disturbance input and frequency response in a power system, specifically including: The input function space, i.e., the space where the disturbance power source resides, is denoted as: ,in, Represents the spatial region of the power grid (such as a geographic coordinate plane); Indicates a time interval; Represent the function space of the connection; The output function space, i.e., the space in which the angular frequency response resides, is denoted as: ; The FNO operator is used to learn a nonlinear operator that, given any perturbation input function, can directly output the corresponding angular frequency response function without relying on a traditional numerical solver. In this way, unlike the finite-dimensional vectors processed by traditional physical networks, this scheme can operate directly in an infinite-dimensional function space, thus not depending on the grid resolution. It can handle inputs with different sampling densities in any space. After training, it can respond to disturbances in the spatial distribution or time length of unknown edges, which is more in line with the dynamic and continuous evolution process of the power system.

[0011] In one feasible implementation, the specific expression of the operator mapping relationship in step 3 includes: ; in, This indicates disturbance input, including power disturbances and fluctuations in renewable energy output; Represents spatial variables, namely the geographical or topological spatial location of the power grid (such as node coordinates or grid points). This represents the angular frequency response, i.e., the spatiotemporal distribution of the angular frequency of a power grid node; Represents the FNO operator; Represents the network parameters of the FNO operator; In this way, by capturing the nonlinear propagation characteristics of the power grid frequency in the spatial (multi-node) and temporal dimensions through Fourier transform, it is possible to quickly approximate the overall evolution law of complex systems. Compared with traditional numerical integration methods, the FNO operator has higher computational efficiency and is suitable for handling power grid systems with hundreds of nodes or even larger scale.

[0012] Step 4: Jointly train the PINN model and the FNO operator to obtain the simulation model.

[0013] In one feasible implementation, the joint training in step 4 includes a joint training loss function, the specific formula of which includes: ; in, Indicates joint loss; The term represents the FNO operator term, which is the error when the FNO operator learns the mapping between the disturbance input (such as power surges, fluctuations in new energy power supply, etc.) and the angular frequency response in the function space. It is used to ensure that the PINN model can efficiently capture the spatiotemporal propagation law in multi-node, large-scale power grids. Indicates the weight coefficient of the FNO operator term; This represents the weighting coefficient of the physical residual term; Indicates the weight coefficient of the boundary condition term; This represents the weight coefficients of the data error term; each weight coefficient can be dynamically adjusted according to the specific stage of joint training to balance the contribution of different loss terms. In this way, through the triple mechanism of "physical constraints + data-driven + operator learning", physical consistency, prediction accuracy and computational efficiency are effectively balanced. During specific training, the PINN model ensures that the solution satisfies the power grid dynamics equations, and the FNO operator provides operator-level fast mapping capability. The two work together to improve prediction accuracy and speed.

[0014] Step 5: Collect the disturbance power value and the random output power value of new energy sources in the power grid under test, input them into the simulation model, predict the angular frequency, and obtain the frequency propagation curve (of multiple nodes in the power grid), so as to quickly predict new disturbance inputs.

[0015] In one feasible implementation, the specific data acquisition method in step 5 includes: By using synchronous phasor measurement units (PMUs) deployed at grid nodes, voltage amplitude and current phasor are measured to derive disturbance power values. The power monitoring system deployed at the new energy power station collects the random output power value of the new energy source. Real-time active power of residential, commercial, and industrial loads is collected through load monitoring terminals (LMD) or smart meter clusters.

[0016] In one feasible implementation, step 5 further includes visualizing the frequency propagation curve for understanding and application.

[0017] Secondly, based on the same inventive concept, this application also provides a power grid frequency propagation simulation system, including a data receiving module, a data processing module and a result generation module; The data receiving module is used to receive the disturbance power value of the power grid under test and the random output power value of new energy sources; The data processing module includes a theoretical model unit, a simulation model unit, and a prediction unit; The theoretical model unit is based on the theory of power grid dynamics and constructs a power grid frequency propagation model. The simulation model unit constructs a PINN model; transforms the power grid dynamics equations into partial differential equation constraints and embeds them into the training process of the PINN model; learns the operator mapping relationship between disturbance input and angular frequency response in the function space through the FNO operator; and jointly trains the PINN model and the FNO operator to obtain the simulation model. The prediction unit inputs the disturbance power value of the power grid under test and the random output power value of new energy into the simulation model to predict the angular frequency and obtain the frequency propagation curve; The result generation module is used to transmit the frequency propagation curve externally.

[0018] Thirdly, based on the same inventive concept, this application also provides a power grid frequency propagation simulation device, including a processor, a memory, and a bus. The memory stores instructions and data read by the processor, and the processor is used to call the instructions and data in the memory to execute the power grid frequency propagation simulation method as described above. The bus connects the various functional components for transmitting information.

[0019] By adopting the above technical solution, the present invention has the following beneficial effects: This invention provides a power grid frequency propagation simulation method, system, and device, which can accurately describe the power grid frequency dynamics under conditions of new energy power supply fluctuations, large disturbances, and complex boundary conditions. By explicitly introducing nonlinear terms and disturbance sources into the model, this scheme can capture complex dynamic processes such as oscillations, frequency drops, and slow recovery, overcoming the limitation of traditional methods that are only applicable to small disturbance conditions. This scheme embeds the power grid dynamics equations (PDE constraints) during model training and uses PINN to ensure the physical consistency of the prediction results. At the same time, it utilizes the operator learning capability of FNO to directly learn the mapping between input and output in the function space, ensuring computational efficiency and scalability. The combination of PINN and FNO avoids the uninterpretability of "black box" models and achieves efficient modeling capabilities. This scheme does not rely on grid partitioning or large-scale numerical integration methods. Instead, it adopts a gridless modeling and operator learning strategy, which can adapt to power grids of different sizes and topologies. In particular, it can maintain high-precision prediction in power grids with hundreds of nodes or even larger scales, thus avoiding the problem of traditional dimensionality reduction modeling failing in complex networks. This solution improves inference speed by about one to two orders of magnitude (10 to 100 times) compared to traditional numerical methods while ensuring prediction error is less than 1%, making it suitable for online application scenarios (such as dispatch control and emergency response). This combination of high precision and high efficiency allows it to meet the needs of scientific research simulation as well as directly serve the operation of the power grid. This solution can be extended to different types of power grid frequency simulation scenarios and is suitable for new energy power grids with different penetration rates. It can also be extended to other dynamic problems of power systems such as voltage stability and transient stability. Even under the condition of scarce training data, this solution can still maintain stable prediction performance, thus demonstrating good versatility and robustness. This solution supports real-time frequency prediction and visualization, helping operators to intuitively grasp the disturbance propagation process. In emergency situations, it can serve as an auxiliary decision-making tool, providing a reference for frequency control and safety protection. Compared with existing methods (which can only be used for offline analysis), this solution is more suitable for the actual needs of power grid dispatching, planning, and new energy grid connection control. Attached Figure Description

[0020] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0021] Figure 1 A flowchart of a power grid frequency propagation simulation method provided in an embodiment of the present invention; Figure 2 The frequency perturbation propagation time-domain curve of this solution provided in the embodiment of the present invention; Figure 3 A comparative illustration of the time-domain curves of frequency disturbance propagation between the present solution and the comparative solution provided in this embodiment of the invention; Figure 4 Color diagram showing the spatial profile distribution of frequency propagation in this embodiment of the invention; Figure 5 This is a screenshot of a dynamic animation showing the propagation of frequency disturbances in this embodiment of the invention. Figure 6 A diagram of a power grid frequency propagation simulation system is provided for an embodiment of the present invention. Detailed Implementation

[0022] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0024] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0025] The present invention will be further explained below with reference to specific embodiments.

[0026] It should also be noted that the specific embodiments or implementation methods described below are a series of optimized settings listed by the present invention to further explain the specific content of the invention, and these settings can be combined or used in conjunction with each other.

[0027] Example 1: like Figure 1 As shown in the figure, the power grid frequency propagation simulation method provided in this embodiment includes the following steps: Step 1: Construct a power grid frequency propagation model based on power grid dynamics theory.

[0028] Furthermore, the power grid frequency propagation model includes power grid dynamics equations, power grid flow constraints, and disturbance terms; The power grid dynamics equation can be a generator swing equation, and the specific calculation formula includes: ; in, Indicates the first The inertial constant of each power grid node; Indicates the first Damping coefficient of each power grid node; Indicates the first The angular frequency of each power grid node; Indicates the rated angular frequency; Indicates the first Mechanical power input of each grid node; Indicates the first Electromagnetic output power of each power grid node; Indicates the total number of nodes in the power grid; Indicates time; The specific expressions for the power flow constraints include: ; in, Indicates the first Power flow values ​​of each grid node; Indicates the first Voltage amplitude at each grid node; Indicates the first Voltage amplitude at each grid node; It represents the conductance of branches between power grid nodes and is used to reflect active power loss; It represents the susceptance of branches between power grid nodes, used to reflect reactive power exchange and line charging effects; Indicates the first Voltage phase angle of each grid node With the Voltage phase angle of each grid node The difference between them, i.e. ; The disturbance term is used to realistically reflect the power grid operating conditions, so as to provide strict PDE constraints and physical priors for the training of subsequent neural networks, thereby ensuring the basic rationality of the model; the specific expression includes: ; in, Indicates the first Mechanical power input values ​​for each grid node; Indicates the first Initial values ​​of mechanical power of each grid node; This indicates the disturbance power value (such as disturbances caused by sudden power changes, generator loss of synchronization, line faults, etc.). This indicates the random output power value of the new energy source.

[0029] Step 2: Construct the PINN model (Physical Information Neural Network); transform the power grid dynamics equations into partial differential equation (PDE) constraints and embed them into the training process of the PINN model.

[0030] Furthermore, the partial differential equation constraints include physical residual terms, boundary condition terms, and data error terms; The physical residual term represents the equation residual after substituting the angular frequency prediction data into the power grid dynamics equations. This ensures that the PINN model output satisfies the physical laws of the power system, thereby avoiding "black box" bias. The specific calculation formula includes: ; in, Represents the physical residual term; Indicates the number of sampling points for the physical residual term; The boundary condition terms are used to constrain the PINN model at the initial time (e.g., The solution at the rated frequency of 50 Hz (or at the boundary node) is used to maintain consistency with actual operating conditions. Specific expressions include: ; in, Represents boundary condition terms; Indicates the number of sampling points for the boundary condition term; This represents the initial value of the angular frequency prediction data; The data error term represents the difference between the predicted angular frequency data and the simulated (or measured) angular frequency data, in order to ensure that the PINN model can effectively fit the simulated (or measured) angular frequency data, thereby improving its adaptability to actual power grid conditions. The specific expression includes: ; in, Indicates the data error term; This represents the angular frequency prediction data; Indicates the first At that moment; Represents simulated (or measured) angular frequency data; This indicates the number of sampling points for the data error term; In this way, variables such as angular frequency, power, time, and grid node location can be used as inputs to the PINN model to predict the dynamic response of the grid node angular frequency. Three types of terms are introduced into the loss function: differential equation residuals, boundary condition residuals, and data errors. This ensures that the PINN model learning process conforms to physical laws and can fit real measurement or simulation data. Compared with pure data-driven methods, the PINN model significantly improves physical consistency and result interpretability.

[0031] Step 3: Using the FNO operator (Fourier neural operator), learn the operator mapping relationship between the perturbation input and the angular frequency response in the function space.

[0032] Furthermore, the specific expression for the operator mapping relationship in step 3 includes: ; in, This indicates disturbance input, including power disturbances and fluctuations in renewable energy output; Represents spatial variables, namely the geographical or topological spatial location of the power grid (such as node coordinates or grid points). This represents the angular frequency response, i.e., the spatiotemporal distribution of the angular frequency of a power grid node; Represents the FNO operator; Represents the network parameters of the FNO operator; The FNO network built based on the FNO operator includes point-by-point multiplication operations in the frequency domain through Fourier transform, and the specific expressions include: ; in, Indicates Fourier transform; Represents (learnable) weights; Indicates the first in the output frequency domain The Fourier coefficients of each angular frequency, i.e., the angular frequency response characteristics after processing by the FNO network; This represents a nonlinear activation function, typically GELU (Gaussian Error Linear Unit) or ReLU; This represents the final angular frequency response, i.e. .

[0033] In this way, by capturing the nonlinear propagation characteristics of the power grid frequency in the spatial (multi-node) and temporal dimensions through Fourier transform, it is possible to quickly approximate the overall evolution law of complex systems. Compared with traditional numerical integration methods, the FNO operator has higher computational efficiency and is suitable for handling power grid systems with hundreds of nodes or even larger scale.

[0034] Step 4: Jointly train the PINN model and the FNO operator to obtain the simulation model.

[0035] Furthermore, the joint training in step 4 includes a joint training loss function, the specific formula of which includes: ; in, Indicates joint loss; The term FNO operator represents the error in which the FNO operator maps perturbation inputs (such as power surges, fluctuations in new energy power supply, etc.) to the angular frequency response in the function space. This error is used to ensure that the PINN model can efficiently capture spatiotemporal propagation patterns in multi-node, large-scale power grids. Indicates the weight coefficient of the FNO operator term; This represents the weighting coefficient of the physical residual term; Indicates the weight coefficient of the boundary condition term; This represents the weight coefficients of the data error term; each weight coefficient can be dynamically adjusted according to the specific stage of joint training to balance the contribution of different loss terms. In this way, through the triple mechanism of "physical constraints + data-driven + operator learning", physical consistency, prediction accuracy and computational efficiency are effectively balanced. During specific training, the PINN model ensures that the solution satisfies the power grid dynamics equations, and the FNO operator provides operator-level fast mapping capability. The two work together to improve prediction accuracy and speed.

[0036] Step 5: Collect the disturbance power value and the random output power value of new energy sources in the power grid under test, input them into the simulation model, predict the angular frequency, and obtain the frequency propagation curve (of multiple nodes in the power grid), so as to quickly predict new disturbance inputs.

[0037] Furthermore, step 5 also includes visualizing the frequency propagation curve for better understanding and application.

[0038] Example 2: A power grid model containing 10 power grid nodes is constructed. Frequency propagation simulation experiments are conducted using the simulation method described in Example 1. Specifically, a disturbance power is applied at power grid node 1, and the spatiotemporal propagation process of the frequency of the entire power grid is simulated. By comparing the traditional numerical method with this scheme, the advantages of this scheme in terms of accuracy and efficiency are verified.

[0039] Experimental setup: Power grid size: 10 power grid nodes; Disturbance method: Apply a power disturbance to the main power line (BUS) at 0.1 seconds; Comparison approach: Numerical integration of traditional difference algebraic equations (DAE); Evaluation metrics: mean square error (MSE), mean absolute error (MAE), root mean square error (RMSE), and simulation time.

[0040] Experimental results: like Figure 2 As shown, after a disturbance occurs, the frequency of each power grid node exhibits a dynamic response over time, characterized by a rapid decline, local oscillations, and a gradual convergence process. The curves for different power grid nodes (Node1-10) demonstrate the temporal pattern of the disturbance propagating from a local location to the entire network. Figure 2 This demonstrates that PINN in this scheme can capture the frequency dynamics of a multi-node power grid while maintaining a high degree of physical consistency.

[0041] like Figure 3 As shown, the prediction curves of this scheme (solid line) and the comparison scheme (dashed line) at each power grid node are in very good agreement, and the error can be controlled at the millihertz (mHz) level. The annotation results show that MSE=0.84mHz and RMSE=1.11mHz, indicating that while ensuring high accuracy, this scheme achieves far greater computational efficiency than traditional numerical methods through AI simulation.

[0042] like Figure 4 As shown, this scheme displays the frequency distribution of each power grid node at the moment of disturbance in a scatter plot, and the frequency values ​​can be represented by color mapping. As time progresses, the spatial diffusion process of the disturbance can also be observed, with the frequencies between nodes deviating sequentially and gradually recovering. This figure verifies that this scheme can not only depict the temporal dynamics but also reveal the spatial propagation characteristics.

[0043] like Figure 5 As shown, this scheme demonstrates the propagation process of disturbances in the power grid; the animation uses time as the evolution axis and presents the frequency response of each power grid node frame by frame, intuitively showing the entire process of disturbance propagation, oscillation of each power grid node, and gradual restoration of system stability; compared with static diagrams, dynamic diagrams are more helpful for operators to understand the disturbance propagation mechanism.

[0044] In summary, in this experiment, the average prediction error of this scheme is less than 1%; the AI's second-level calculation speed is about 10 to 50 times faster than traditional numerical simulation; and it can still maintain stable prediction performance under conditions of new energy fluctuations and large disturbances.

[0045] Example 3: like Figure 6 As shown, this embodiment provides a power grid frequency propagation simulation system, including a data receiving module, a data processing module, and a result generation module; The data receiving module is used to receive the disturbance power value of the power grid under test and the random output power value of new energy sources; The data processing module includes a theoretical model unit, a simulation model unit, and a prediction unit; The theoretical model unit is based on the theory of power grid dynamics and constructs a power grid frequency propagation model. The simulation model unit constructs a PINN model; transforms the power grid dynamics equations into partial differential equation constraints and embeds them into the training process of the PINN model; learns the operator mapping relationship between disturbance input and angular frequency response in the function space through the FNO operator; and jointly trains the PINN model and the FNO operator to obtain the simulation model. The prediction unit inputs the disturbance power value of the power grid under test and the random output power value of new energy into the simulation model to predict the angular frequency and obtain the frequency propagation curve; The result generation module is used to transmit the frequency propagation curve externally.

[0046] Example 4: This embodiment provides a power grid frequency propagation simulation device, including a processor, a memory, and a bus. The memory stores instructions and data read by the processor, and the processor is used to call the instructions and data in the memory to execute the power grid frequency propagation simulation method described above. The bus connects the various functional components for information transmission.

[0047] In another embodiment, this solution can also be implemented using an integrated device, which may include corresponding modules that perform one or more steps in the various embodiments described above. A module may be one or more hardware modules specifically configured to perform the corresponding step, or implemented by a processor configured to perform the corresponding step, or stored in a computer-readable medium for implementation by a processor, or implemented through some combination thereof.

[0048] The processor executes the various methods and processes described above. For example, the method implementations in this scheme can be implemented as software programs tangibly contained in a machine-readable medium, such as memory. In some implementations, part or all of the software program can be loaded and / or installed via memory and / or a communication interface. When the software program is loaded into memory and executed by the processor, one or more steps of the methods described above can be performed. Alternatively, in other implementations, the processor can be configured to execute one of the methods described above by any other suitable means (e.g., by means of firmware).

[0049] This device can be implemented using a bus architecture. A bus architecture can include any number of interconnect buses and bridges, depending on the specific application of the hardware and overall design constraints. The bus connects various circuits, including one or more processors, memory, and / or hardware modules. The bus can also connect various other circuits such as peripherals, voltage regulators, power management circuitry, external antennas, etc.

[0050] Buses can be Industry Standard Architecture (ISA) buses, Peripheral Component Interconnect (PCI) buses, or Extended Industry Standard Component (EISA) buses, etc. Buses can be divided into address buses, data buses, control buses, etc.

[0051] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for simulating power grid frequency propagation, characterized in that, include: Step 1: Construct a power grid frequency propagation model based on power grid dynamics theory; Step 2: Construct the PINN model; transform the power grid dynamics equations into partial differential equation constraints and embed them into the training process of the PINN model; Step 3: Using the FNO operator, learn the operator mapping relationship between the perturbation input and the angular frequency response in the function space; Step 4: Jointly train the PINN model and the FNO operator to obtain the simulation model; Step 5: Collect the disturbance power value of the power grid under test and the random output power value of new energy sources, input them into the simulation model, predict the angular frequency, and obtain the frequency propagation curve.

2. The simulation method according to claim 1, characterized in that, The power grid frequency propagation model includes power grid dynamics equations, power flow constraints, and disturbance terms.

3. The simulation method according to claim 2, characterized in that, The power grid dynamics equation is the generator swing equation, and the specific calculation formula includes: ； in, Indicates the first The inertial constant of each power grid node; Indicates the first Damping coefficient of each power grid node; Indicates the first The angular frequency of each power grid node; Indicates the rated angular frequency; Indicates the first Mechanical power input of each grid node; Indicates the first Electromagnetic output power of each power grid node; Indicates the total number of nodes in the power grid; Indicates time.

4. The simulation method according to claim 2, characterized in that, The specific expressions for the power flow constraints include: ； in, Indicates the first Power flow values ​​of each grid node; Indicates the first Voltage amplitude at each grid node; Indicates the first Voltage amplitude at each grid node; Indicates the electrical conductance of branches between power grid nodes; Indicates the susceptance of branches between power grid nodes; Indicates the first Voltage phase angle of each grid node With the Voltage phase angle of each grid node The difference between them.

5. The simulation method according to claim 2, characterized in that, The perturbation term has the following specific expression: ； in, Indicates the first Mechanical power input values ​​for each grid node; Indicates the first Initial values ​​of mechanical power of each grid node; Indicates the disturbance power value; This indicates the random output power value of the new energy source.

6. The simulation method according to claim 3, characterized in that, The partial differential equation constraints include physical residual terms, boundary condition terms, and data error terms; The specific calculation formula for the physical residual term includes: ； in, Represents the physical residual term; Indicates the number of sampling points for the physical residual term; The boundary condition term specifically includes the following expressions: ； in, Represents boundary condition terms; Indicates the number of sampling points for the boundary condition term; This represents the initial value of the angular frequency prediction data; The data error term has the following specific expression: ； in, Indicates the data error term; This represents the angular frequency prediction data; Indicates the first At that moment; This represents the measured angular frequency data; This indicates the number of sampling points for the data error term.

7. The simulation method according to claim 2, characterized in that, The specific expression for the operator mapping relationship in step 3 includes: ； in, Indicates a disturbance input; Represents spatial variables; Indicates the angular frequency response; Represents the FNO operator; This represents the network parameters of the FNO operator.

8. The simulation method according to claim 6, characterized in that, Step 4, the joint training, includes a joint training loss function, the specific formula of which is: ； in, Indicates joint loss; Represents the FNO operator term; Indicates the weight coefficient of the FNO operator term; This represents the weighting coefficient of the physical residual term; Indicates the weight coefficient of the boundary condition term; This represents the weighting coefficient for the data error term.

9. A power grid frequency propagation simulation system, characterized in that, It includes a data receiving module, a data processing module, and a result generation module; The data receiving module is used to receive the disturbance power value of the power grid under test and the random output power value of new energy sources; The data processing module includes a theoretical model unit, a simulation model unit, and a prediction unit; The theoretical model unit is based on the theory of power grid dynamics and constructs a power grid frequency propagation model. The simulation model unit constructs a PINN model; transforms the power grid dynamics equations into partial differential equation constraints and embeds them into the training process of the PINN model; learns the operator mapping relationship between disturbance input and angular frequency response in the function space through the FNO operator; and jointly trains the PINN model and the FNO operator to obtain the simulation model. The prediction unit inputs the disturbance power value of the power grid under test and the random output power value of new energy into the simulation model to predict the angular frequency and obtain the frequency propagation curve; The result generation module is used to transmit the frequency propagation curve externally.

10. A power grid frequency propagation simulation device, characterized in that, It includes a processor, a memory, and a bus. The memory stores instructions and data read by the processor. The processor is used to call the instructions and data in the memory to execute the simulation method as described in any one of claims 1-8. The bus connects the functional components for transmitting information.

Images