Self-adaptive multi-resolution power system simulation method and device based on physical information

Abstract

The invention provides a self-adaptive multi-resolution power system simulation method and device based on physical information, and relates to the technical field of power system simulation. The method comprises the following steps: acquiring model data and operation state data of a power system; constructing a simulation system corresponding to the power system based on the model data, the operation state data and the multi-level resolution model library; determining a target model resolution corresponding to each subsystem through a self-adaptive resolution control strategy and a multi-level resolution model library; and distributing the target model resolution to the corresponding subsystem, and performing simulation calculation on the subsystem of the simulation system based on the multi-level resolution model library and the adaptive resolution control strategy to obtain a simulation calculation result corresponding to the power system. According to the embodiment of the invention, by designing the adaptive resolution control strategy, the dynamic adjustment of the resolution of different regional models of the power system is realized, and the calculation efficiency can be greatly improved while the simulation precision is maintained.

Description

Technical Field

[0001] The present invention relates to the technical field of power system simulation, and in particular to a physical information-based adaptive multi-resolution power system simulation method and device. Background Art

[0002] As power systems decarbonize, the proportion of renewable energy sources such as wind and solar connected to the grid via power electronic converters continues to increase. These converters introduce fast switching dynamics, ranging from microseconds to milliseconds, which coexist with slower electromechanical oscillations and control interactions, ranging from milliseconds to seconds / minutes, resulting in complex multi-timescale dynamic behavior. Accurate simulation is crucial for assessing power system stability, designing control strategies, and ensuring reliable operation.

[0003] Traditional simulation methods face a dilemma. Electromagnetic transient (EMT) simulations, which use detailed models and small time steps (e.g., 1-50 seconds), can capture fast dynamics with high accuracy, but their computational burden becomes unmanageable for large-scale systems containing thousands of buses and hundreds of renewable energy units. Conversely, root mean square (RMS) simulations, which use simplified models and larger time steps (e.g., 1-10 milliseconds), are suitable for electromechanical phenomena but ignore the fast converter dynamics that are critical for evaluating short-circuit behavior, subsynchronous oscillations, or control interactions.

[0004] Hybrid EMT-RMS simulation attempts to bridge this gap by modeling critical areas (e.g., renewable energy plants, HVDC stations) in EMT detail, while simulating the rest of the system in RMS form. However, existing hybrid approaches often rely on a predefined fixed partitioning. This lacks adaptability: a disturbance occurring in an RMS region may require a detailed EMT analysis, or a detailed region may become quiet, allowing computational savings through model simplification.

[0005] Therefore, in the process of power system simulation, how to adaptively adjust the model resolution is a technical problem that needs to be solved urgently. Summary of the Invention

[0006] The present invention provides a physical information-based adaptive multi-resolution power system simulation method and device to address the existing technical issues of the contradiction between computational efficiency and accuracy in the simulation of large-scale power systems containing high-penetration renewable energy. By designing an adaptive resolution control strategy, dynamic adjustment of the model resolution of different areas of the power system is achieved, which can significantly improve computational efficiency while maintaining simulation accuracy.

[0007] In a first aspect, the present invention provides a method for adaptive multi-resolution power system simulation based on physical information, comprising the following steps: Obtain model data and operating status data of the power system; Building a simulation system corresponding to the power system based on the model data, the operating status data and a multi-resolution model library, wherein the simulation system includes at least a plurality of different subsystems; Determining the target model resolution corresponding to each subsystem through the adaptive resolution control strategy and the multi-level resolution model library; The target model resolution is allocated to the corresponding subsystem, and simulation calculations are performed on the subsystems of the simulation system based on a multi-level resolution model library and an adaptive resolution control strategy to obtain simulation calculation results corresponding to the power system.

[0008] Preferably, according to a physical information-based adaptive multi-resolution power system simulation method provided by the present invention, the multi-resolution model library includes at least an electromagnetic transient model, a detailed root mean square model, a simplified root mean square model, a physical information neural network dynamic equivalent model, a quasi-steady-state model, and a fixed load model; The resolution level definition of the adaptive resolution control strategy includes at least: a first accuracy level corresponding to the electromagnetic transient model, a second accuracy level corresponding to the detailed root mean square model, a third accuracy level corresponding to the simplified root mean square model, a fourth accuracy level corresponding to the physical information neural network dynamic equivalent model, and a fifth accuracy level corresponding to the quasi-steady-state model and / or the fixed load model; The first level of accuracy is characterized by the use of an electromagnetic transient model with a time step of 1-50s to capture switching transients, harmonics, and fast control dynamics; The second accuracy level is characterized by the use of a detailed RMS model with a time step of 1-5 ms to capture electromechanical oscillations and detailed renewable energy control dynamics; The third level of accuracy is characterized by using a simplified RMS model with a time step of 5-10ms to capture frequency dynamics and initial voltage stability; The fourth level of accuracy is characterized by the use of a physical information neural network dynamic equivalent model with a time step greater than 1ms, which is used to represent aggregate dynamic behavior and learn complex interactions; The fifth accuracy level is characterized by using a quasi-steady-state model and / or a fixed load model, a time step greater than 100 ms, and only minimal dynamic characteristics.

[0009] Preferably, according to the physical information-based adaptive multi-resolution power system simulation method provided by the present invention, the target model resolution corresponding to each subsystem is determined by the adaptive resolution control strategy and the multi-level resolution model library, including: Based on the multi-criteria triggering mechanism in the adaptive resolution control strategy and the multi-level resolution model library, the target model resolution corresponding to each subsystem is determined; wherein, the target model resolution is one of the first accuracy level, the second accuracy level, the third accuracy level, the fourth accuracy level and the fifth accuracy level determined based on the multi-criteria triggering mechanism.

[0010] Preferably, according to the physical information-based adaptive multi-resolution power system simulation method provided by the present invention, the multi-criteria triggering mechanism in the adaptive resolution control strategy includes: An event-based trigger, wherein the event-based trigger is used to respond to discrete events detected by the monitoring module of the simulation system, wherein the discrete events include at least faults, topology changes, control mode switching, and protection actions; A state-based trigger, wherein the state-based trigger is used to simulate system variables by continuously monitoring the system, wherein the system variables include at least voltage deviation, frequency deviation, power oscillation detection and key control signal activity; a sensitivity-based trigger for evaluating a potential impact on a region of interest, wherein the potential impact includes at least an impact on the region of interest and a model difference metric; A stillness detection trigger is used to detect minimum dynamic activity of a high-resolution subsystem within a duration when the resolution is reduced.

[0011] Preferably, according to the physical information-based adaptive multi-resolution power system simulation method provided by the present invention, after the step of assigning the target model resolution to the corresponding subsystem, the method includes: Perform smooth resolution switching with state variable mapping to achieve seamless transitions between multi-resolution models; The step of performing smooth resolution switching with state variable mapping comprises: In the conversion between the root mean square model and the electromagnetic transient model, simulation continuity is ensured by mapping between phasors and instantaneous values; wherein the root mean square model is one of the detailed root mean square model and the simplified root mean square model; In the conversion between different levels of RMS models, the common state is transferred and the remaining state variables are processed; In the conversion between the root mean square model and the physical information neural network dynamic equivalent model, the state is initialized based on the input relationship or output relationship of the neural network; Use smooth transitions or projection methods to handle discontinuities and minimize numerical transients.

[0012] Preferably, according to the physical information-based adaptive multi-resolution power system simulation method provided by the present invention, the allocating the target model resolution to the corresponding subsystem includes: Dynamically allocate target model resolution to different subsystems based on real-time system events, state deviations, and sensitivity analysis; The dynamic allocation of target model resolutions to different subsystems based on real-time system events, state deviations, and sensitivity analysis also includes: A hybrid impact domain identification strategy based on multi-index dynamic evaluation is used to determine the disturbance impact area; The hybrid impact domain identification strategy includes: Initial rapid delineation stage: combining topological information with simplified electrical distances to quickly determine the initial candidate impact domain within a very short time after the disturbance occurs; Dynamic boundary monitoring and adjustment phase: Dynamically expand or shrink the impact domain by real-time monitoring of multiple indicators such as the state deviation gradient, power flow change rate, local frequency change rate, and control signal activity of boundary nodes and neighboring nodes; Stabilization and regional management stage: When the main dynamic process of the disturbance tends to ease, the impact domain is finally stabilized based on the mechanism of impact domain stability confirmation, impact domain merging and final shrinkage / elimination of the impact domain.

[0013] Preferably, according to the physical information-based adaptive multi-resolution power system simulation method provided by the present invention, the process of shrinking the influence domain in the dynamic boundary monitoring and adjustment stage includes: Define the core impact area: Identify the area around the initial incident source or the area containing critical equipment as the core area that will not participate in the contraction; Identify candidate shrinkage areas: Identify components / subsystems that are inside the candidate influence domain but outside the core influence area as candidate shrinkage areas; Evaluating the dynamics of the candidate contraction region: monitoring local static indicators of the components within the candidate contraction region, wherein the local static indicators include at least: whether the rate of change of the state variable is lower than a threshold, whether the deviation from the quasi-steady-state value is lower than a threshold, and whether the relevant control signal is inactive; Evaluate the connectivity of candidate shrinkage regions: Check whether removing a candidate shrinkage region would fragment the remaining high-resolution regions or sever the connection between the core influence region and the current active boundary; Perform shrinkage: When the candidate shrinkage area meets the local static conditions, does not belong to the core influence area, and its removal does not destroy the necessary connectivity, the corresponding resolution is reduced and removed from the current high-resolution influence domain.

[0014] Preferably, according to the physical information-based adaptive multi-resolution power system simulation method provided by the present invention, the step of constructing the physical information neural network dynamic equivalent model includes: Offline training is performed using high-precision simulation data, and physical constraints are incorporated into the loss function during training. Predict the subsystem's response based on boundary conditions online to provide a dynamic representation of the simulated system; The switching between the physical information neural network dynamic equivalent model and the remaining resolution level models is achieved through an adaptive resolution control strategy.

[0015] Preferably, according to the physical information-based adaptive multi-resolution power system simulation method provided by the present invention, the simulation calculation steps include: Initialization processing: setting the initial time, initial state and initial resolution; Monitoring and processing: obtaining current status and detecting events; Evaluation triggers: Identify subsystems that require resolution changes based on a multi-criteria trigger mechanism; Determine the target resolution: For each subsystem that needs to change the resolution, determine the new resolution level; Switch resolution: Pause the subsystem whose resolution needs to be switched, obtain the current state, load the new multi-resolution model, map the state variables, update the interface and resume the simulation; Advance simulation: Solve the simulation system equations using the current resolution configuration, obtain the solution, and update the state based on the solution; Repeat the above monitoring process to advance the simulation steps until the simulation ends.

[0016] In a second aspect, the present invention further provides a physical information-based adaptive multi-resolution power system simulation device, comprising: an acquisition module, configured to acquire model data and operating status data of a power system, wherein the power system is a power system containing a high penetration rate of renewable energy; a construction module, configured to construct a simulation system corresponding to the power system based on the model data, the operating status data, and a multi-resolution model library, wherein the simulation system includes at least a plurality of different subsystems; A determination module, configured to determine a target model resolution corresponding to each subsystem by using an adaptive resolution control strategy and the multi-level resolution model library; The simulation calculation module is used to allocate the target model resolution to the corresponding subsystem, and perform simulation calculation on the subsystem of the simulation system based on the multi-level resolution model library and the adaptive resolution control strategy to obtain simulation calculation results corresponding to the power system.

[0017] In a third aspect, the present invention also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements any one of the physical information-based adaptive multi-resolution power system simulation methods described above.

[0018] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the physical information-based adaptive multi-resolution power system simulation methods described above.

[0019] In a fifth aspect, the present invention further provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the physical information-based adaptive multi-resolution power system simulation methods described above.

[0020] The present invention provides a method and device for simulating a power system based on adaptive multi-resolution physical information. The method and device obtain the model data and operating status data of the power system; construct a simulation system corresponding to the power system based on the model data, the operating status data and the multi-level resolution model library, wherein the simulation system includes at least a plurality of different subsystems; determine the target model resolution corresponding to each subsystem through an adaptive resolution control strategy and the multi-level resolution model library; assign the target model resolution to the corresponding subsystem, and perform simulation calculations on the subsystems of the simulation system based on the multi-level resolution model library and the adaptive resolution control strategy to obtain simulation calculation results corresponding to the power system. The method is used to solve the contradiction between computational efficiency and accuracy in the simulation of large-scale power systems containing high-penetration renewable energy in the prior art. By designing an adaptive resolution control strategy, dynamic adjustment of the model resolution of different regions of the power system is achieved, which can significantly improve computational efficiency while maintaining simulation accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0022] Figure 1 This is one of the flow charts of the physical information-based adaptive multi-resolution power system simulation method provided by the present invention.

[0023] FIG2 is a schematic diagram of a physical information adaptive multi-resolution simulation framework provided by the present invention.

[0024] FIG3 is a flow chart of the adaptive resolution control strategy provided by the present invention.

[0025] FIG4 is a schematic diagram of multi-resolution model conversion provided by the present invention.

[0026] FIG5 is a schematic diagram of a dynamic equivalent model of a physical information neural network provided by the present invention.

[0027] FIG6 is a schematic diagram showing the comparison between simulation accuracy and computational efficiency provided by the present invention.

[0028] FIG7 is a schematic diagram of the structure of the physical information-based adaptive multi-resolution power system simulation device provided by the present invention.

[0029] Figure 8 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0030] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0031] In the related art, there are at least the following technical problems: Inspired by adaptive mesh refinement in computational fluid dynamics and level-of-detail techniques in computer graphics, the concept of dynamically adjusting model resolution during simulation holds significant potential for power systems. Furthermore, recent advances in data-driven modeling, particularly physically-informed neural networks (PINNs), offer a promising avenue for creating computationally efficient and physically plausible dynamic equivalent models. PINNs embed physical laws into the training process, potentially overcoming the generalization and interpretability limitations of purely black-box models.

[0032] To address the limitations of fixed-resolution approaches and capitalize on these advances, a solution is needed that can dynamically adjust model accuracy based on system conditions during simulation, while maintaining the ability to accurately capture critical dynamics.

[0033] The following combination Figures 1-8The present invention describes a physical information-based adaptive multi-resolution power system simulation method and device, which is used to address the contradiction between computational efficiency and accuracy in the simulation of large-scale power systems containing high penetration of renewable energy in the existing technology. By designing an adaptive resolution control strategy, dynamic adjustment of the model resolution of different areas of the power system is achieved, which can significantly improve computational efficiency while maintaining simulation accuracy.

[0034] Figure 1 This is one of the flow charts of a method for adaptive multi-resolution power system simulation based on physical information provided by the present invention, such as Figure 1 As shown, the method may include but is not limited to steps S100 to S400: S100, obtaining model data and operating status data of the power system; S200, constructing a simulation system corresponding to the power system based on the model data, the operating status data, and a multi-resolution model library, wherein the simulation system includes at least a plurality of different subsystems; S300, determining the target model resolution corresponding to each subsystem through the adaptive resolution control strategy and the multi-level resolution model library; S400, allocating the target model resolution to the corresponding subsystem, and performing simulation calculations on the subsystems of the simulation system based on a multi-level resolution model library and an adaptive resolution control strategy to obtain simulation calculation results corresponding to the power system.

[0035] In step S100 of some embodiments, model data and operating status data of a power system are acquired, where the power system is a power system containing high penetration rate of renewable energy.

[0036] The acquired model data may at least include but is not limited to power system component model parameters, network topology, and load model.

[0037] The model parameters of power system components at least include: for traditional power generation equipment such as thermal power units and hydropower units, their model parameters usually include the rated capacity, maximum ramp rate, minimum head (for hydropower units), etc. of the generator.

[0038] For renewable energy generation equipment such as wind turbines and photovoltaic arrays, in addition to basic parameters such as rated power and rated voltage, unique characteristic parameters also need to be considered. For example, the wind turbine's wind power utilization coefficient curve (Cp curve) describes the efficiency of the wind turbine in converting wind energy into mechanical energy at different wind speeds; the photovoltaic array's light intensity-power curve reflects the power generated by photovoltaic panels under different lighting conditions.

[0039] The network topology structure includes: node information and branch information.

[0040] Node information: Determines the location, number, and connectivity of each node in the power system (e.g., power plants, substations, load centers, etc.). Node information can be obtained from power system geographic information system (GIS) data, grid planning drawings, and actual dispatch management systems.

[0041] Branch information: This includes parameters such as impedance, admittance, and transformation ratio of branch components such as transmission lines and transformers. These parameters can be determined through nameplate markings on power equipment, experimental test results, and calculations of the power system's short-circuit current.

[0042] Load model includes: load characteristic analysis and load modeling method.

[0043] Load characteristics analysis: Through statistical analysis of historical load data, we can understand the load variation patterns in different seasons and time periods, such as daily load curves and annual load curves. At the same time, we should consider the components of the load, such as the proportional relationship between industrial load, commercial load, and residential load.

[0044] Load Modeling Methods: Common load modeling methods include dynamic load models based on physical mechanisms, static load models based on data statistics, and comprehensive load models that combine the advantages of both. Select the appropriate modeling method based on specific needs and identify and verify model parameters based on actual data.

[0045] The operating status data of the power system can be obtained through real-time monitoring systems and communication networks.

[0046] In step S200 of some embodiments, a simulation system corresponding to the power system is constructed based on the model data, the operating status data, and a multi-resolution model library, and the simulation system includes at least a plurality of different subsystems.

[0047] First, the model data and operating status data collected in step S100 are sorted, cleaned, and formatted to ensure data accuracy and consistency. According to the requirements of the multi-level resolution model library, the data is classified and layered to provide a basis for subsequent modeling.

[0048] Establish a multi-level resolution model library: The multi-level resolution model library at least includes electromagnetic transient model, detailed root mean square model, simplified root mean square model, physical information neural network dynamic equivalent model, quasi-steady-state model, and fixed load model.

[0049] The resolution level definition of the adaptive resolution control strategy includes at least: a first accuracy level corresponding to the electromagnetic transient model, a second accuracy level corresponding to the detailed root mean square model, a third accuracy level corresponding to the simplified root mean square model, a fourth accuracy level corresponding to the physical information neural network dynamic equivalent model, and a fifth accuracy level corresponding to the quasi-steady-state model and / or the fixed load model.

[0050] The first accuracy level is represented as high accuracy level, the second accuracy level is represented as medium-high accuracy level, the third accuracy level is represented as medium accuracy level, the fourth accuracy level is represented as low accuracy level, and the fifth accuracy level is represented as the lowest accuracy level.

[0051] The first level of accuracy is characterized by the use of an electromagnetic transient model with a time step of 1-50s to capture switching transients, harmonics and fast control dynamics.

[0052] Specifically, the high precision level (first precision level) may at least include the following steps: 1) Component-level fine modeling: For power generation equipment, whether traditional or renewable energy generators, we model them down to their most fundamental components. For example, we build precise dynamic models for each component of a wind turbine, including its blades, gearbox, and generator, taking into account detailed parameters such as mass, moment of inertia, and friction coefficient, as well as their interactions and energy conversion processes. For photovoltaic cells, we build models based on semiconductor physics to accurately simulate processes such as the photovoltaic effect and carrier transport.

[0053] In terms of load modeling, various types of loads are decomposed into basic combinations of components such as resistance, inductance, and capacitance. The parameters of each component are determined according to the type and characteristics of the load to accurately simulate the changes in the voltage and current characteristics of the load under different working conditions.

[0054] Grid components such as transformers and transmission lines also use detailed distributed parameter models to fully consider influencing factors such as skin effect and proximity effect in high-frequency transient processes, and accurately describe the electromagnetic transient processes of the power grid.

[0055] 2) Accurate description of microscopic topological structure: Detailed micro-topological connectivity of each subsystem within the power system is constructed, including the connection method of each winding within the generator and the specific electrical connection between the load and the grid access point. Accurately simulate the signal transmission path and time delay between components to ensure accurate capture of rapidly changing electromagnetic transient processes and local dynamic responses in the system.

[0056] 3) Accurate simulation of complex control strategies: High-precision mathematical models are developed for generator control systems, such as maximum power point tracking (MPPT) and low voltage ride-through control in wind power generation, as well as traditional generator control strategies such as automatic speed regulation and automatic excitation regulation. These models accurately reflect the dynamic adjustment process of control strategies under different operating conditions and fault conditions, and precisely simulate the real-time regulation of power generation equipment by the control system.

[0057] 4) Advanced data processing and analysis: In terms of data acquisition, high-speed, high-precision data acquisition equipment is used to monitor the operating status of the power system in real time, collecting various signals including voltage, current, power, frequency, wind speed, solar radiation intensity, etc. The sampling frequency is high enough to capture rapidly changing transient processes.

[0058] Advanced data processing algorithms are used to accurately analyze and process the collected data, extracting key features and information for model validation, state assessment, and prediction. For example, methods such as wavelet transform are used to analyze non-stationary signals to accurately identify transient responses and changing trends in the system.

[0059] For automated control systems in power grids, such as automatic voltage control (AVC) and automatic generation control (AGC), accurate models are also established to consider various factors in the generation, transmission, and execution of control instructions, such as communication delay and control cycle, in order to achieve precise control of the power grid's operating status.

[0060] The second level of accuracy is characterized by the use of detailed RMS models with a time step of 1-5 ms to capture electromechanical oscillations and detailed renewable energy control dynamics.

[0061] Specifically, the medium-high precision level (second precision level) includes the following steps: 1) Equipment-level aggregate modeling: Modeling power generation equipment as individual units, moving beyond the component-level internal structure to focus on the overall electrical and mechanical characteristics of the unit. For example, for a wind turbine, a comprehensive wind speed-power characteristic model and pitch angle control model are developed. For traditional generators, a transfer function model is developed to describe the dynamic characteristics of the speed control system and reheat system.

[0062] Load modeling adopts a comprehensive load model to aggregate loads of the same type (such as industrial loads, commercial loads, residential loads, etc.), and establish corresponding models based on their overall power factor, voltage characteristics, etc., without distinguishing the parameters of each load component in detail.

[0063] 2) Topological modeling combining macro and micro: When describing the topology of a power system, we consider both the overall macroscopic grid structure and, to a certain extent, retain the microscopic connection information of key nodes. For example, a more detailed model is used for the internal electrical connections of a large power plant, while a simplified equivalent circuit model is used for the connection between the power plant and the external grid. This reduces model complexity while maintaining a certain level of accuracy.

[0064] 3) Control strategy simplification and equivalence: Appropriate simplifications can be made to generator control strategies. For example, maximum power point tracking control in wind power generation can be simplified to a power setting model based on average wind speed, and the complex speed regulation control of traditional generators can be equivalent to a PID control model with typical parameters. Automatic control systems in power grids can also adopt simplified models, such as simplifying automatic voltage control (AVC) to a control model based on regional reactive power balance to reduce computational complexity.

[0065] 4) Data processing and analysis optimization: In terms of data acquisition, the sampling frequency is appropriately reduced while still ensuring that the system's key dynamic processes are captured. Data compression and filtering techniques are used to remove noise and redundant information, improving data processing efficiency. In terms of data analysis, statistical analysis methods and simple dynamic models are used to process data and quickly assess the system's status and performance.

[0066] The third accuracy level is characterized by using a simplified RMS model with a time step of 5-10ms to capture frequency dynamics and initial voltage stability.

[0067] Specifically, the medium precision level (third precision level) may at least include but not be limited to the following steps: 1) Unit-level modeling: Modeling takes the generator set as the basic unit, eliminating detailed consideration of the internal microstructure and component characteristics. Instead, the entire unit is considered as a whole to describe its electrical and mechanical characteristics. For example, for wind turbines, empirical models based on measured data are developed, such as power curve models and wind speed-power correlation models. For traditional generators, simplified second-order rotor swing equations are used to describe their mechanical dynamic characteristics.

[0068] Load modeling uses typical industry load curves and power factor data to establish a simplified load model. Different types of loads are combined in a certain proportion to form a comprehensive load model to reflect the overall change trend of the load.

[0069] 2) Macro topology modeling: The focus is on describing the macroscopic grid structure of the power system, appropriately aggregating and simplifying grid components such as substations and transmission lines. For example, short-distance, low-voltage transmission lines are combined into a single equivalent line, and small substations are simplified to voltage conversion nodes, thus significantly reducing the complexity of the grid model.

[0070] 3) Basic control function simulation: Generator control functions are simplified for simulation. For example, wind turbines only consider basic speed control and pitch angle limiting, while traditional generators only simulate simple speed regulation and excitation control. Automatic control systems in power grids use simplified logic models. For example, automatic voltage control (AVC) only considers proportional-integral control based on simple voltage deviations, while automatic generation control (AGC) uses a simple regulation strategy based on frequency deviations.

[0071] 4) Data processing and statistical evaluation: Data acquisition primarily focuses on key system operating parameters, such as voltage, current, power, and frequency. While the sampling frequency is further reduced, it still captures the system's steady state and slowly changing dynamics. For data processing, simple statistical analysis methods and trend analysis tools are used to process the collected data and assess the system's overall performance and stability.

[0072] The fourth level of accuracy is characterized by the use of a physical information neural network dynamic equivalent model with a time step greater than 1ms, which is used to represent aggregate dynamic behavior and learn complex interactions.

[0073] Specifically, the low-precision level (fourth precision level) may at least include but not be limited to the following steps: 1) Single-machine equivalent modeling: Simplify an entire power plant or large-scale renewable energy generation cluster into an equivalent generator model. For traditional power plants, an equivalent synchronous generator model is established based on parameters such as total installed capacity and rated voltage. For renewable energy generation clusters, such as wind farms or photovoltaic power plants, a simplified equivalent power source model is established based on their overall power characteristics and grid access characteristics.

[0074] Load modeling uses a very simplified constant power load model or an approximate model based on a typical load curve, without considering the dynamic changes in the load details, and only focusing on the overall power consumption level.

[0075] 2) Simplify topology modeling: This approach greatly simplifies the power system topology, retaining only the main grid skeleton structure, such as high-voltage transmission lines and large substations. It reduces the complex power grid network to a few nodes and simple branch connections, completely ignoring the microscopic electrical connections and distributed parameter characteristics of grid components.

[0076] 3) Application of basic physical laws: During the modeling process, only basic physical laws and simple linear relationships are considered, such as Ohm's law and Kirchhoff's law. Complex electromagnetic transient processes, nonlinear characteristics, and the influence of distributed parameters are ignored, and a very simple mathematical model is used to describe the operating state of the power system.

[0077] 4) Data estimation and approximate processing: Data collection relies primarily on statistical data and empirical estimates, with a low sampling frequency, providing only a rough estimate of some basic system operating parameters. Data processing involves simple arithmetic operations and logical reasoning, with minimal requirements for data accuracy. The primary goal is to provide a general understanding of system operation.

[0078] The fifth accuracy level is characterized by using a quasi-steady-state model and / or a fixed load model, a time step greater than 100 ms, and only minimal dynamic characteristics.

[0079] Specifically, the lowest accuracy level (fifth accuracy level) may at least include but is not limited to the following steps: 1) Overall estimation modeling: This approach treats the entire power system as a black box, modeling it solely based on the relationship between its inputs (such as total power input power and total load power) and outputs (such as the overall power supply of the grid). This approach completely ignores the system's internal structure and operating mechanisms, using the simplest mathematical expressions or empirical formulas to describe the system's input-output relationship.

[0080] 2) High-level overview of the topological structure: It does not involve any actual grid topology details, but only uses a simple schematic diagram or symbol to represent the existence of the power system, without the concept of specific nodes, branches or components.

[0081] 3) Idealized assumptions and simplifications: A large number of idealized assumptions are made, such as that the power supply is completely stable, the load is constant, and the transmission of the power grid is lossless and delay-free. The various complex factors and uncertainties in the actual system are not considered, and the operation of the power system is simplified to an ideal, static process.

[0082] 4) Minimal data processing: There is almost no data collection, or only some total data is recorded, such as the total power generation and total power consumption of the day. Data processing is limited to simple comparison and sorting operations, without in-depth data analysis and processing algorithms.

[0083] In some embodiments of the present invention, as shown in FIG2 , the physical information adaptive multi-resolution simulation framework provided by the present invention includes several key components: The framework consists of a multi-resolution model library, a simulation engine core, a monitoring module, an adaptive resolution controller, a model management module, and a data exchange interface. The framework aims to strike a balance between simulation accuracy and computational efficiency by dynamically adapting to the modeling details of different parts of the power system.

[0084] In some embodiments of the present invention, a model library containing multiple resolution levels is constructed. As shown in Table 1, these resolution levels are represented by a resolution level definition table.

[0085] Table 1 is as follows:

[0086] When determining the applicability of each resolution level, the EMT (Electromagnetic Transient Model) model is typically applied to a single converter or small electrical island; the RMS (Root Mean Square) model is typically applied to larger network areas; and the PINN (Physical Information Neural Network) dynamic equivalent model can represent anything from a single complex device to an entire regional network. The Adaptive Resolution Controller (ARCS) uses this information to select the appropriate target resolution during switching, among the EMT model, detailed RMS model, simplified RMS model, PINN dynamic equivalent model, quasi-steady-state model, and fixed load model.

[0087] In some embodiments, the partitioning and modeling of subsystems: Generator subsystem: Based on different power generation types (conventional power generation and renewable energy power generation) and modeling accuracy requirements, various generator models are established using corresponding model data, including their control systems, speed regulation systems, etc., to simulate the output characteristics and operating behavior of the generators.

[0088] Load subsystem: Considering the type, characteristics and changing rules of the load, combined with the load model data, a load model is established, including static load and dynamic load, to reflect the impact of the load on the power system.

[0089] Grid subsystem: Based on the network topology and component parameters, a grid model is established, including transformers, transmission lines, busbars, etc., to simulate the transmission and distribution process of electric energy.

[0090] Control and protection subsystem: Establish control strategy models for the power system, such as automatic generation control (AGC), automatic voltage control (AVC), and protection device models to ensure system stability and safety.

[0091] Connect each subsystem through interfaces to form a complete power system simulation system, ensuring that each subsystem can interact and work together correctly.

[0092] The simulation system is verified using historical operation data and actual operation conditions, the differences between simulation results and actual conditions are compared, the model parameters and system structure are adjusted and optimized, and the accuracy and reliability of the simulation system are improved.

[0093] In step S300 of some embodiments, the target model resolution corresponding to each subsystem is determined through the adaptive resolution control strategy and the multi-level resolution model library.

[0094] Furthermore, the target model resolution corresponding to each subsystem is determined by the adaptive resolution control strategy and the multi-level resolution model library, including: Based on the multi-criteria triggering mechanism in the adaptive resolution control strategy and the multi-level resolution model library, the target model resolution corresponding to each subsystem is determined; wherein, the target model resolution is one of the first accuracy level, the second accuracy level, the third accuracy level, the fourth accuracy level and the fifth accuracy level determined based on the multi-criteria triggering mechanism.

[0095] Furthermore, the adaptive resolution control strategy is a control method that automatically adjusts the resolution according to different situations. In power system simulation, it allows the system to determine the appropriate model resolution based on actual needs and operating conditions, balancing computational efficiency and simulation accuracy.

[0096] The multi-resolution model library includes a series of models from the first level of accuracy (high precision) to the fifth level of accuracy (low precision). These models have different levels of complexity and accuracy to meet different application scenarios and analysis requirements.

[0097] A multi-criteria triggering mechanism considers multiple criteria or standards to assess when to switch model resolution. These criteria might include the speed of the system's dynamic changes, the occurrence of faults, specific operating conditions, and more. By combining these criteria, it can more intelligently determine whether to adjust the model resolution to suit the current situation.

[0098] The target model resolution is determined based on a multi-criteria trigger mechanism. The system selects the model resolution that best matches the current state of each subsystem from a library of multi-resolution models. This resolution may be the first level (high precision), used for detailed analysis and accurate simulation; or the second, third, fourth, or fifth levels (decreasing in accuracy), used for scenarios with less stringent accuracy requirements.

[0099] By combining this adaptive resolution control strategy with a multi-resolution model library, power system simulation can flexibly switch between different accuracy levels, ensuring simulation and analysis at the appropriate resolution under various operating conditions. This not only improves computational efficiency but also provides high-precision simulation results during critical periods, supporting decision-making and system optimization.

[0100] In some embodiments of the present invention, the multi-criteria triggering mechanism in the adaptive resolution control strategy includes: An event-based trigger, wherein the event-based trigger is used to respond to discrete events detected by the monitoring module of the simulation system, wherein the discrete events include at least faults, topology changes, control mode switching, and protection actions; A state-based trigger, wherein the state-based trigger is used to simulate system variables by continuously monitoring the system, wherein the system variables include at least voltage deviation, frequency deviation, power oscillation detection and key control signal activity; a sensitivity-based trigger for evaluating a potential impact on a region of interest, wherein the potential impact includes at least an impact on the region of interest and a model difference metric; A stillness detection trigger is used to detect minimum dynamic activity of a high-resolution subsystem within a duration when the resolution is reduced.

[0101] It is understandable that if Figure 3 As shown, the Adaptive Resolution Control Strategy (ARCS) is a core component of this invention. It uses a multi-criteria triggering mechanism to determine when and how to change the resolution level of a subsystem. The system is initialized and performs real-time status monitoring and event detection. When an event is triggered based on the multi-criteria triggering mechanism, the subsystem that needs to be changed is determined, and the target resolution is determined based on the adaptive resolution control strategy. Resolution switching is performed based on the target resolution, thereby advancing the simulation time step. If the multi-criteria triggering mechanism does not trigger a condition, the current resolution is maintained to advance the simulation time step.

[0102] These triggering mechanisms include: multi-criteria triggering mechanism. The multi-criteria triggering mechanism includes the following steps: 1) Event-based triggers: The discrete events that respond include at least: Faults: When a short circuit occurs near a subsystem, it immediately triggers an increase in resolution. For example, upgrading to the higher-accuracy L1 / EMT level is achieved by monitoring circuit breaker status, line current / voltage, and other factors to determine if a fault has occurred.

[0103] Topology changes: For example, line or generator tripping. Topology changes are detected by monitoring switch states, which can trigger a higher resolution (for example, focusing on more detailed electrical connection changes) or a lower resolution (if the changed topology is relatively simple).

[0104] Control mode switching: For example, when a renewable energy converter changes modes (e.g., from grid-following mode to grid-forming mode, or performing fault ride-through), it triggers an increase in resolution, such as upgrading to L1 / L2. This can be achieved by monitoring control flags or inferring control mode switching from terminal behavior.

[0105] Protection action: Triggered when a relay operates. You can monitor the status of the analog relay or infer relay action based on relevant thresholds to determine whether resolution changes are necessary.

[0106] 2) State-based triggers: The monitored system variables and corresponding operations include at least: Voltage / Frequency Deviation: Calculates voltage deviation ( )、Frequency Deviation( ), voltage rate of change (|dV / dt|), and frequency rate of change (|df / dt|). When these values ​​exceed the set threshold, it is recommended to increase the resolution. For example, when increasing from level L3 to level L2, the threshold check is used to determine whether the conditions for increasing the resolution are met.

[0107] Power Oscillation Detection: Triggers when a significant oscillation is detected that does not decay or grow. Signal processing techniques such as the Brownie method, FFT (Fast Fourier Transform), or matrix brushstrokes are used to determine whether to trigger a resolution increase, such as from L3 to L2 or L1.

[0108] Control signal activity: Monitors critical control signals for high variance or saturation. By monitoring signal statistics, such as variance, we can determine if higher precision is needed to more accurately describe the system state, thereby deciding whether to increase resolution from L3 to L2 or L1 levels.

[0109] 3) Sensitivity-based triggers (predictive / proactive) The potential impacts and methods of assessment shall at least include: Impact on the Area of ​​Interest (AOI): Estimates the sensitivity of AOI variables to subsystem states. This evaluation is performed using pre-calculated or updated sensitivities and information such as electrical distance. If sensitivity is high and the subsystem is experiencing disturbances, the resolution is increased.

[0110] Model Difference Metrics: Compare the low-resolution model to the expected physical or detected high-resolution model to check for significant deviations. Define an error metric to determine if a change in resolution is warranted.

[0111] 4) Inactivity detection trigger (for reducing resolution) Detection Condition and Action: Triggered when the high-resolution subsystem displays minimal dynamic activity for a duration (Tstable). This condition is determined by monitoring state derivatives / deviations, allowing a switch to a lower resolution, such as from L1 to L2, or from L2 to L3 or L4.

[0112] These triggers are combined using logic rules in an adaptive resolution control strategy (ARCS) to form a decision-making mechanism that automatically adjusts the resolution level of the subsystem according to different system conditions, thereby improving computational efficiency while ensuring simulation accuracy.

[0113] In some embodiments of the present invention, after the step of assigning the target model resolution to the corresponding subsystem, the method includes: Perform smooth resolution switching with state variable mapping to achieve seamless transitions between multi-resolution models; The step of performing smooth resolution switching with state variable mapping comprises: In the conversion between the root mean square model and the electromagnetic transient model, simulation continuity is ensured by mapping between phasors and instantaneous values; wherein the root mean square model is one of the detailed root mean square model and the simplified root mean square model; In the conversion between different levels of RMS models, the common state is transferred and the remaining state variables are processed; In the conversion between the root mean square model and the dynamic equivalent model of the physical information neural network, the state is initialized based on the input relationship or output relationship of the neural network; Use smooth transition or projection methods to handle discontinuities in smooth resolution switching and minimize numerical transients.

[0114] It's understandable that after assigning the target model resolution to the corresponding subsystem, smooth resolution switching with state variable mapping is required to achieve seamless transitions between multi-resolution models. This is because models of different resolutions differ in aspects such as state variable representation. Direct switching can lead to simulation interruptions or inaccuracies, while smooth resolution switching ensures simulation continuity and accuracy.

[0115] As shown in Figure 4, performing a resolution switch involves selecting the target resolution, loading a new model, and key state variable mapping. State variable mapping ensures simulation continuity and specifically includes the following parts: The operations and meanings of different types of conversions are as follows: 1) Conversion between RMS model and electromagnetic transient model Mapping between phasors and instantaneous values: When converting between L2 / L3 (RMS) and L1 (EMT), a mapping between phasors and instantaneous values ​​is required. Sinusoidal states are typically assumed and interface variables are matched, and the controller states need to be mapped / reinitialized. This is to ensure that the representation of variables such as voltage is correctly converted when converting between two models with different resolutions, and that the operating states of the controller are correctly connected. For voltage variables, the formula for converting from root mean square (RMS) representation to instantaneous value is: (1) in, is the voltage amplitude, is the phase angle, is the rotational speed.

[0116] This formula realizes the conversion from RMS value to instantaneous value, taking into account factors such as the phase relationship of the three-phase voltage.

[0117] 2) Conversion between different levels of RMS models Transferring common states and processing remaining state variables: When converting between different levels of RMS models (such as L2 and L3), common states are transferred and other states are discarded, approximated, or initialized. Taking the generator as an example, examples of mapping from a detailed RMS model to a simplified RMS model include: (2) (3) (4) Where δ is the power angle, ω is the speed, L2 represents the second accuracy level corresponding to the detailed RMS model, and L3 represents the third accuracy level corresponding to the simplified RMS model.

[0118] These mapping relationships ensure that key state variables (such as power angle δ, speed ω, etc.) can be correctly transferred and processed when converting between RMS models of different resolutions, so that the simulation can continue.

[0119] 3) Conversion between the RMS model and the physical information neural network dynamic equivalent model State initialization based on the neural network's input or output relationships: When transitioning between L2 / L3 (RMS) and L4 (PINN-DE), depending on the PINN's input / output relationships, warm-up or optimization may be required to initialize the PINN state. During the reverse transition, the PINN output provides boundary conditions for the RMS model. This ensures that the model's state is correctly initialized and that physical consistency is maintained during model transitions involving the neural network.

[0120] In some embodiments, methods for handling discontinuities include using smooth transitions or projections to handle discontinuities during resolution switching to minimize numerical transients. This is to ensure that during the model resolution switching process, there are no unnecessary fluctuations or errors in the simulation results due to sudden changes in state variables.

[0121] In some embodiments, the steps and key elements for performing resolution switching include: 1. Selecting a target resolution, loading a new model, and mapping key state variables: Based on the simulation requirements and system status, select an appropriate target resolution and load the corresponding new model. At the same time, determine the mapping relationship of key state variables, which is key to ensuring simulation continuity.

[0122] 2. Specific operations of state variable mapping to ensure simulation continuity: L2 / L3 (RMS) L1 (EMT) mapping: As mentioned above, mapping between phasors and instantaneous values, as well as mapping / reinitialization of controller states, etc.

[0123] L2L3 (RMS level) mapping: transfer common states, discard / approximate / initialize other states, such as the mapping example of generator-related state variables.

[0124] L2 / L3 (RMS) L4 (PINN - DE) mapping: Perform state initialization and provide boundary conditions according to the characteristics of PINN.

[0125] Dealing with discontinuities: Using smooth transitions or projection methods to reduce numerical transient issues caused by discontinuities.

[0126] In some embodiments of the present invention, allocating the target model resolution to the corresponding subsystem includes: Dynamically assign target model resolution to different subsystems based on real-time system events, state deviations, and sensitivity analysis.

[0127] Furthermore, dynamic allocation is performed based on multiple factors: model resolution is dynamically assigned to target subsystems based on real-time system events, state deviations, and sensitivity analysis. This means that the system flexibly adjusts the resolution of each subsystem to better suit the system's operating status and simulation requirements based on different situations and needs.

[0128] The dynamic allocation of target model resolutions to different subsystems based on real-time system events, state deviations, and sensitivity analysis also includes: A hybrid impact domain identification strategy based on multi-index dynamic evaluation is used to determine the disturbance impact area; The hybrid impact domain identification strategy includes: Initial rapid delineation stage: combining topological information with simplified electrical distances to quickly determine the initial candidate impact domain within a very short time after the disturbance occurs; Dynamic boundary monitoring and adjustment phase: Dynamically expand or shrink the impact domain by real-time monitoring of multiple indicators such as the state deviation gradient, power flow change rate, local frequency change rate, and control signal activity of boundary nodes and neighboring nodes; Stabilization and regional management stage: When the main dynamic process of the disturbance tends to ease, the impact domain is finally stabilized based on the mechanism of impact domain stability confirmation, impact domain merging and final shrinkage / elimination of the impact domain.

[0129] In some embodiments of the present invention, a novel Hybrid Dynamic Influence Zone Identification (HD-IZI) strategy based on dynamic multi-metric evaluation is proposed to determine the spatial extent of subsystems requiring resolution change. This strategy is divided into three stages: 1. Initial Rapid Demarcation (IRD) phase. 2. Dynamic Boundary Monitoring & Adjustment (DBMA) phase. 3. Stabilization & Zone Management (SZM) phase.

[0130] Phase 1: Initial Rapid Delineation (IRD) Objective: To quickly identify the initial candidate influence zone (CIZ) within a very short time (within a few simulation steps) after a disturbance occurs. This is primarily triggered by strong events (such as faults and high-power equipment switching).

[0131] The specific steps include: Event source location: Accurately determine the location where the event occurred, such as faulty lines, tripped equipment, etc.

[0132] Combining topology and reduced electrical distance: Starting from the event source, perform N - order topological proximity search (N is selected according to the event type, e.g., for a fault N = 2, for a line trip N = 1).

[0133] Simultaneously calculate or query the reduced electrical distance metrics between the event source and neighboring nodes.

[0134] Set the combined threshold: Topological hop count ≤ N and reduced electrical distance < Dthresh. The innovation in this stage is to quickly combine topology and electrical distance information to preliminarily screen the relevant area. The reduced electrical distance between each pair of nodes or the list of "electrical neighbors" can be pre - calculated offline, and only table - lookup and simple topological search are required online. All components within the output CIZ area will immediately be upgraded to a preset higher resolution.

[0135] Stage 2: Dynamic Boundary Monitoring and Adjustment Stage (DBMA) Objective: After the initial CIZ is determined, continuously monitor the dynamic behavior of its boundary and the neighboring areas, and dynamically expand or contract the affected area according to the actual impact propagation. This stage runs continuously during the existence of the CIZ or is activated by a status trigger.

[0136] Specific steps are as follows: Definition of boundary nodes: Identify the nodes directly connected inside and outside the CIZ (boundary nodes) and the nodes directly connected to the boundary nodes outside the CIZ (neighboring nodes).

[0137] Multi - metric monitoring: State deviation gradient: Calculate the difference in state variables such as voltage, frequency, phase angle, etc. (or the difference in the rate of change) between the nodes on both sides of the boundary, and it may be normalized using line impedance, e.g., , a large gradient indicates that the impact crosses the boundary.

[0138] Rate of change of energy / power flow: Monitor the rate of change of active / reactive power of the lines crossing the boundary. A rapid and continuous change in power flow indicates the impact propagation. ​​​​​​​Dynamic adjustment logic: Expansion: If one or more indicators of a neighboring node (outside the CIZ) exceed the preset expansion threshold, it is included in the influence domain and the resolution is improved.

[0142] Contraction: Define the core impact zone (CoIZ), which is generally the area around the initial incident source or containing critical equipment, and is usually not involved in contraction.

[0143] Identify candidate constriction zones (CCZs), i.e., elements / subsystems inside the CIZ but outside the CoIZ.

[0144] Evaluate CCZ dynamics and monitor its local static indicators, such as the rate of change of state variables below the threshold (|dx / dt|< x), the deviation from the quasi-steady-state value is less than the threshold value (|x - xss_ref|< dev), and related control signals are inactive or unsaturated; assess CCZ connectivity and examine whether removing the CCZ will fragment the remaining high-resolution area or sever the connection between the CoIZ and the active boundary.

[0145] Contraction: When a CCZ satisfies local static conditions, is not part of a CoIZ, and its removal does not disrupt essential connectivity, its resolution is reduced and removed from the high-resolution influence domain. This phase not only relies on boundary information but also incorporates dynamic evaluation of the internal region and connectivity analysis, enabling more intelligent region contraction, reclaiming computational resources while avoiding impacting critical dynamic simulations.

[0146] Phase 3: Stabilization and Zone Management (SZM) Objective: After the main dynamic process of the disturbance has eased, the impact domain is stabilized and multiple independent impact domains are managed. This is triggered when the frequency of expansion and contraction activities in the DBMA phase decreases significantly, or when the overall dynamic indicators of the system tend to stabilize. Specific steps: Influence domain stability confirmation: If the boundary of an influence domain does not expand or shrink significantly within the set time window Tstable_IZ, and the internal dynamics are below the stability threshold, the influence domain is considered to be relatively stable.

[0147] Influence domain merging: In complex scenarios (such as cascading failures), monitor the interactions between independent influence domains, such as power oscillations on cross-regional lines and voltage / phase angle fluctuations between boundary nodes. If the interaction indicator continuously exceeds the merge threshold (Thresholdmerge), it indicates strong coupling dynamics and should be merged into a larger influence domain.

[0148] Final shrinkage / elimination of the influence domain: When the system enters a new stable state or quasi-steady state, the resolution of the high-resolution influence domain is gradually reduced to the background resolution based on the global stationary detection logic.

[0149] Key points of HD-IZI strategy technology implementation: Data structure: Use graph databases or adjacency lists to efficiently represent system topology, the resolution of each subsystem, the boundary nodes of the influence domain, and other information.

[0150] Threshold setting: Based on offline simulation analysis and system characteristic setting, adaptive threshold or machine learning-assisted methods can also be considered.

[0151] Computational efficiency: The computation of the IRD phase needs to be completed within a few time steps.

[0152] The amount of monitoring computation in the DBMA phase is proportional to the boundary length, and some of the computations can be parallelized.

[0153] Robustness: Introduce hysteresis characteristics or time delay confirmation to solve the "jitter" problem, and handle the fallback mechanism when model switching fails or numerical instability occurs.

[0154] Advantages of HD-IZI strategy: Hybridity: Combining static (such as topology information) and dynamic (such as real-time state deviation, power flow changes, etc.) information.

[0155] Dynamic adaptation: Dynamically evolve the impact domain based on the actual propagation of the disturbance.

[0156] Multi-stage refinement: from initial rapid rough delineation to subsequent fine dynamic adjustment.

[0157] Smart contraction: Safe contraction areas are determined based on internal dynamics and connectivity analysis.

[0158] Scenario adaptability: Able to handle disturbances of different types and severity.

[0159] In some embodiments of the present invention, the process of shrinking the influence domain in the dynamic boundary monitoring and adjustment stage includes: Define the core impact area: Identify the area around the initial incident source or the area containing critical equipment as the core area that will not participate in the contraction; Identify candidate shrinkage areas: Identify components / subsystems that are inside the candidate influence domain but outside the core influence area as candidate shrinkage areas; Evaluating the dynamics of the candidate contraction region: monitoring local static indicators of the components within the candidate contraction region, wherein the local static indicators include at least: whether the rate of change of the state variable is lower than a threshold, whether the deviation from the quasi-steady-state value is lower than a threshold, and whether the relevant control signal is inactive; Evaluate the connectivity of candidate shrinkage regions: Check whether removing a candidate shrinkage region would fragment the remaining high-resolution regions or sever the connection between the core influence region and the current active boundary; Perform shrinkage: When the candidate shrinkage area meets the local static conditions, does not belong to the core influence area, and its removal does not destroy the necessary connectivity, the corresponding resolution is reduced and removed from the current high-resolution influence domain.

[0160] It is understood that the steps to define the core impact area include: Identify critical areas: Define the area surrounding the initial event source or containing critical equipment as the core area that is not involved in the contraction. This core impact area remains relatively stable throughout the process and is unaffected by the contraction operation. The goal is to ensure that important areas are always in a high-resolution state to accurately capture key dynamic information. For example, in a power system, the area where critical equipment such as generators are located may be defined as the core impact area because changes in state here are critical to the stability and operation of the entire system.

[0161] The steps to identify candidate contraction regions include: Determine the location of candidate regions: Components or subsystems within the candidate impact domain but outside the core impact zone are identified as candidate shrinkage zones. These areas are potential targets for shrinkage operations and require further evaluation to determine whether they meet the shrinkage criteria. For example, in the power system described above, some load areas far from the generator may be initially classified as part of the candidate impact domain. However, if subsequent evaluation determines that their resolution can be reduced, they may become candidate shrinkage zones.

[0162] The steps for evaluating the dynamics of a candidate contraction zone include: Monitor local static indicators, including the rate of change of state variables: Check whether the rate of change of state variables (such as voltage and frequency) of components within the candidate contraction region is below a set threshold. If the state variables change slowly, the region is relatively stable, and high resolution may not be necessary. For example, if the rate of change of voltage in a region remains below the threshold for a long period of time, this indicates minimal voltage fluctuations, and you may consider reducing the resolution in that region.

[0163] Deviation from Quasi-Steady-State Value: Determines whether the deviation of the component state from the quasi-steady-state value is less than a threshold. If the deviation is small, the component state is close to stable, and the need for high-resolution monitoring in this area is reduced. For example, if the deviation of the current of a transmission line from the quasi-steady-state value is consistently less than a threshold, it indicates that the current state is stable, and the resolution can be appropriately reduced.

[0164] Relevant control signal activity: Observe whether relevant control signals (such as those for automatic voltage regulators and automatic speed controllers) within the candidate contraction zone are inactive. Inactive control signals indicate that dynamic regulation requirements in this zone are low, and resolution reduction may be considered. For example, if the automatic voltage regulator control signal in a certain zone remains inactive for an extended period, this indicates low voltage regulation requirements in this zone, and resolution reduction may be appropriate.

[0165] The steps to evaluate the connectivity of a candidate contraction region include: Check the connectivity of the remaining high-resolution areas: Ensure that the removal of the candidate shrinkage area does not disrupt the remaining high-resolution areas. Fragmentation of the remaining high-resolution areas can lead to inaccurate system analysis or loss of effective monitoring of critical components. For example, in a power grid model, if shrinking a region results in a loss of electrical connectivity between two important high-resolution monitoring areas, this can affect the overall assessment of the grid's status.

[0166] Checking connections to core influence areas: Ensure that removing a candidate shrinkage area does not sever the connection between the core influence area and the currently active boundary. This connection between the core influence area and the active boundary is crucial for accurate monitoring and analysis of system dynamics and cannot be disrupted by shrinkage operations. For example, in a distributed energy system, core generation areas exchange power and control with surrounding active load areas, and shrinkage operations must not disrupt this connection.

[0167] The steps to perform a shrink are: After the conditions are met, a contraction operation is performed: When the candidate contraction area simultaneously meets the local static conditions (low rate of change of state variables, small deviation from quasi-steady-state values, and inactive related control signals), does not belong to the core influence area, and its removal does not destroy the necessary connectivity, the resolution corresponding to the area is reduced and removed from the current high-resolution influence domain. Through such an operation, the computing resource usage can be reasonably reduced while ensuring the monitoring accuracy of key parts of the system, thereby improving the efficiency of simulation or monitoring. For example, in a large-scale power system simulation, by contracting the area that meets the conditions, the number of nodes and the amount of calculation can be reduced without affecting the accurate monitoring of key equipment and active boundaries.

[0168] In some embodiments of the present invention, the step of constructing the physical information neural network dynamic equivalent model includes: Offline training is performed using high-precision simulation data, and physical constraints are incorporated into the loss function during training. Predict the subsystem's response based on boundary conditions online to provide a dynamic representation of the simulated system; The switching between the physical information neural network dynamic equivalent model and the remaining resolution level models is achieved through an adaptive resolution control strategy.

[0169] It is understandable that if Figure 5 As shown, the Physical Information Neural Network (PINN) dynamic equivalent model (L4) builds a bridge between the detailed physical model and the simplified representation: Role: Efficiently represent complex subsystems while preserving key dynamic characteristics, suitable for active areas that do not require the highest accuracy; Physical information properties: Incorporating physical laws into the training loss helps generalization and interpretability; Training: Offline training using high-precision simulation data, guided by physics; Reasoning: Online prediction of subsystem responses based on boundary conditions; Adaptability: Can switch between higher / lower resolution levels (L1 / L2 / L3 / L4 / L5) via ARCS triggers and state mapping.

[0170] During PINN training, the loss function includes data matching terms and physical constraint terms: (5) in, Measuring the difference between the model output and the training data, measures the residual of the physical equation, and λ is the weight coefficient used to balance the relative importance of data matching terms and physical constraint terms.

[0171] For power systems, physical constraints can include power balance equations, generator swing equations, etc. (6) Where, Expressed as a physical constraint, Expressed as a power balance equation, Expressed as the generator swing equation, M is the generator inertia time constant, ω is the generator angular velocity, is the mechanical power, is electromagnetic power) etc. as physical constraints.

[0172] Furthermore, the steps for determining the dynamic equivalent model of the physical information neural network include: 1. Offline training (using high-precision simulation data and incorporating physical law constraints) Data preparation for offline training: Collect high-precision simulation data that accurately reflects the system's various operating states and characteristics. For power systems, this may include detailed data such as node voltage, branch power, and generator output under different operating conditions.

[0173] This embodiment ensures the diversity and representativeness of the data, covering various situations that the system may encounter, such as different load levels, failure scenarios, etc.

[0174] Incorporating physical laws into the loss function: During the training process, physical laws are incorporated into the loss function in the form of constraints. For example, in the power system, the power balance equation (node ​​injection power equals node outflow power plus node loss power, i.e. =0) and the generator swing equation ( =( ), where M is the generator inertia time constant, ω is the generator angular velocity, is the mechanical power, is electromagnetic power) etc. as physical constraints.

[0175] The loss function is in the form of ,in Used to measure the difference between model output and training data, It is used to measure the residual of the physical equation, and λ is the weight coefficient used to balance the relative importance of data matching terms and physical constraints. By minimizing this loss function, the model can learn data features while satisfying physical laws.

[0176] 2. The steps of online prediction (predicting subsystem response based on boundary conditions) include: Receive boundary condition input: In actual operation, obtain the system boundary condition information in real time, such as the dispatch instructions issued by the upper power grid in the power system (such as active power setting value, reactive power setting value, etc.), real-time changes in load (such as sudden increase or decrease in load), etc.

[0177] Inference based on a trained model: Utilizing the PINN dynamic equivalent model obtained through offline training, the system can rapidly predict subsystem responses based on input boundary conditions. For example, this can predict voltage changes and power flows at a substation or transmission line under different boundary conditions. Because the model has learned the system's physical laws and data characteristics, it can provide reasonable predictions in a short period of time, thus supporting the dynamic representation of the simulation system.

[0178] 3. Steps of adaptive resolution control strategy (to achieve model switching at different resolution levels): ARCS triggers and state mapping mechanisms: ARCS triggers and state mapping mechanisms are established to determine when switching between higher and lower resolution levels is necessary. ARCS triggers can detect when the system is in a state requiring a resolution switch based on key system metrics, such as the rate of change of state variables and deviation. For example, when the rate of change of voltage in a region exceeds a certain threshold, a switch from a low-resolution model to a high-resolution model is triggered.

[0179] State mapping is responsible for transferring and converting necessary information between models of different resolutions, ensuring a smooth and accurate switching process. For example, when switching from a low-resolution model to a high-resolution model, key information from the low-resolution model (such as voltage and power at major nodes) is transferred to the high-resolution model based on the state mapping relationship, serving as initial conditions or reference information for the high-resolution model.

[0180] Resolution level switching example (L1 / L2 / L3 / L4 / L5): Assume there are five resolution levels (L1-L5), with L1 having the highest resolution and L5 the lowest. When the system is operating normally and state changes are slow, models with lower resolution levels (such as L3 or L4) can be used for calculations and analysis to improve computational efficiency. When the system experiences a major disturbance or a dramatic state change, ARCS triggers detect the relevant conditions and automatically switch to a higher-resolution model (such as L2 or L1) for more precise analysis. For example, when a short circuit occurs in the power system, a detailed simulation requires switching to a high-resolution model (such as L1) to accurately analyze the fault transient process and develop appropriate control strategies.

[0181] 4. The functions and advantages of the PINN dynamic equivalent model are as follows: Role: Effectively represent complex subsystems while preserving key dynamic characteristics, suitable for areas where the highest accuracy is not required. In a large power system, local subsystems that are far from key areas of interest or have little impact on the overall system can be simplified using the PINN dynamic equivalent model, reducing computational effort while still reflecting their dynamic behavior to a certain extent.

[0182] Physical Information: Incorporating physical laws into the training loss improves generalization and interpretability. By considering physical laws, the model not only learns the surface characteristics of the data but also deeply understands the physical nature of the system, enabling better generalization under varying operating conditions. For example, under new, unseen operating conditions, the model can make reasonable predictions based on the laws of physics. Furthermore, the constraints based on physical laws make the model's predictions more interpretable, enabling analysis of the system's operating status and changing trends from a physical perspective.

[0183] Adaptability: ARCS triggers and state mapping can be used to switch between higher and lower resolution levels (L1 / L2 / L3 / L4 / L5). This adaptability allows the model to flexibly adjust computational accuracy and resource allocation based on the real-time needs of the system, improving computational efficiency while maintaining a certain level of accuracy. For example, during daily operations, a lower-resolution model can be used for rapid monitoring and preliminary analysis, and when problems or special cases are discovered, a higher-resolution model can be switched for in-depth analysis.

[0184] In some embodiments of the present invention, the simulation calculation step includes: Initialization processing: setting the initial time, initial state and initial resolution; Monitoring and processing: obtaining current status and detecting events; Evaluation triggers: Identify subsystems that require resolution changes based on a multi-criteria trigger mechanism; Determine the target resolution: For each subsystem that needs to change the resolution, determine the new resolution level; Switch resolution: Pause the subsystem whose resolution needs to be switched, obtain the current state, load the new multi-resolution model, map the state variables, update the interface and resume the simulation; Advance simulation: Solve the simulation system equations using the current resolution configuration, obtain the solution, and update the state based on the solution; Repeat the above monitoring process to advance the simulation steps until the simulation ends.

[0185] It is understood that, as shown in FIG6 , the simulation algorithm process of the present invention includes: 1. Initialization processing: Set the initial time, initial state, and initial resolution: Initial time (t= ): Determines the time at which the simulation begins. This serves as a starting reference point, from which all simulation calculations proceed. For example, in a power system simulation, the initial time might be set to the system startup time or the start time of a specific operation (such as a fault occurrence or load change).

[0186] Initial state ( ): Defines the system's operating state at the initial time. For a power system, this may include the initial values ​​of parameters such as the voltage amplitude and phase angle at each node, the generator power angle and speed, and the load power. These initial states should accurately reflect the system's actual starting conditions and form the basis for subsequent simulation calculations.

[0187] Initial resolution ( ): Sets the initial resolution level of the system. Resolution determines the level of detail in the model's description of the system. For example, in a power grid model, a higher resolution might mean a finer division of grid nodes and branches, enabling a more accurate simulation of power system operation. A lower resolution, on the other hand, might simplify some details, making it suitable for rapid assessment of overall system performance or preliminary analysis of larger systems.

[0188] 2. Monitoring and processing: Obtaining the current state x(t): At each monitoring moment during the simulation, various system status information is collected. This includes the operating parameters of each system component (such as generators, transformers, and loads). For example, in power system simulation, it is necessary to obtain data such as the current active and reactive power output of the generator, and the voltage and current on each side of the transformer to understand the system's operating status in real time.

[0189] This status information is obtained through sensors, monitoring devices, or by directly extracting data from simulation models. This data will serve as an important basis for subsequent evaluation triggers and simulation advancement.

[0190] Detection Events E(t): Simultaneously, the system monitors whether specific events have occurred. Events can include, but are not limited to, faults (e.g., line shorts, line breaks, generator failures), load changes (e.g., sudden increases or decreases in load), and operational events (e.g., switch opening and closing operations, transformer tap adjustments, etc.).

[0191] Detected events provide critical information for determining whether resolution switching is necessary and how to adjust the simulation. For example, if a short circuit fault is detected, it may be necessary to switch to a higher-resolution model in the fault area to more accurately simulate the fault transient process and assess its impact on the system.

[0192] 3. Evaluation trigger: Identify subsystems that require resolution changes based on a multi-criteria trigger mechanism ( ): Establish a multi-criteria evaluation system to determine which subsystems require resolution changes. These criteria may involve multiple aspects, such as the rate of change of subsystem state variables, the amount of change, and whether the subsystem is close to a critical state.

[0193] For example, for a power subsystem, if the rate of change of its node voltage exceeds a preset threshold, or the voltage amplitude deviation exceeds the normal range and the duration exceeds a certain limit, then it may be considered that the subsystem needs to change the resolution and add it to the In collection.

[0194] This multi-criteria evaluation method can comprehensively consider the impact of different factors on subsystems, more accurately identify subsystems that require resolution adjustment, avoid unnecessary resolution switching, and ensure timely response to status changes of key subsystems.

[0195] 4. Determine the target resolution: For each subsystem s that needs to be changed: determine the new resolution level ( (s)): Select an appropriate new resolution level for the subsystem based on its characteristics and current state. If the subsystem is experiencing rapidly changing or complex operating conditions (such as a power subsystem during a fault transient), the resolution may need to be increased to more accurately simulate its dynamic behavior. For relatively stable subsystems with minimal impact on the overall system, the resolution can be appropriately reduced to improve computational efficiency.

[0196] For example, when a power system fault occurs, the resolution of subsystems near the fault area may be switched from a lower level to a higher level to more accurately simulate the fault propagation and equipment response process; while for subsystems far away from the fault area, if their state changes are small, the original resolution can be maintained or appropriately reduced.

[0197] 5. Switch resolution: Pause the subsystem s that require resolution switching: Temporarily pause the simulation calculation of subsystem s before switching resolutions. This ensures data consistency and accuracy during the resolution switching process. Continuing the simulation during the resolution switching process may cause confusion and inaccuracy in the state data, affecting the reliability of subsequent simulation results.

[0198] For example, in power system simulation, if you want to switch the resolution of a subsystem where a generator is located, you need to pause the relevant calculations of the generator before switching to prevent data conflicts or errors during the switching process.

[0199] Get the current state x(t): Record the state information of the subsystem s at the time of pause. This state information includes the operating parameters of each component in the subsystem, such as voltage, current, and power.

[0200] These state data will serve as an important basis for reinitializing the subsystem after loading the new model, ensuring that the subsystem can continue simulation calculations from the correct state at the new resolution.

[0201] Loading a new model (s): According to the new resolution level determined (s), load the corresponding new resolution model for subsystem s. The new model may have different structures, parameters, and calculation methods to adapt to the new resolution requirements.

[0202] For example, when increasing the resolution, it may be necessary to load a more detailed device model to consider more electrical parameters and dynamic characteristics; when reducing the resolution, a simplified equivalent model can be used to increase the calculation speed.

[0203] Mapping Status : Subsystem s at original resolution Current status Map to new resolution This is a critical step because models of different resolutions may describe the system differently, and a reasonable mapping method is needed to convert the original state into the corresponding value in the new state.

[0204] Mapping methods may involve mathematical transformations, interpolation algorithms, and other technical means. For example, in a power system model, it may be necessary to interpolate and calculate the values ​​of state variables such as voltage and current at new nodes based on the node division at the new resolution.

[0205] Update R(t)← : Update the resolution of subsystem s to the new resolution level , and recorded in the system's resolution configuration information R(t). This means that the subsystem will now perform subsequent simulation calculations according to the new resolution.

[0206] At the same time, the interface information associated with that subsystem must be updated to ensure that data interaction and connections with other subsystems remain valid at the new resolution. For example, in a power system, if the resolution of the lines or transformers connected to a substation changes, the electrical connection interface parameters between the substation and other components must be updated accordingly.

[0207] Resume s: After completing the above preparations, restart the simulation calculation of subsystem s. At this time, subsystem s will continue to participate in the simulation process of the entire system at the new resolution based on the updated state and model.

[0208] 6. Advancing Simulation: Solving the System Equations Using the Current Resolution Configuration R(t) (Multiple rates are possible): Configure R(t) based on the current system resolution and solve for the time step The system equations in the power system describe the electrical relationships and dynamic behaviors between the various components in the power system, such as the rotor swing equation of the generator and the power balance equation of the load.

[0209] Solving the system equation is one of the core steps of simulation calculation. By solving it, we can obtain the state change of the system in the next time step.

[0210] In some cases, a multi-rate solution approach may be employed. For example, different time steps can be used to solve for components in a power system with different dynamic characteristics (such as fast-changing power electronics and relatively slow-changing generators) to improve computational efficiency and accuracy.

[0211] Update state x(t+ ): According to the results of the system equations obtained by solving, the system is updated at the next moment t+ The state variable x(t+ ). This includes updating parameters such as the voltage of each node, the current and power of each component, etc.

[0212] The updated state will be used as the initial condition for the simulation calculation of the next time step, allowing the simulation to continue to advance.

[0213] 7. Repeat step: time advances t←t+ : Increase the current time t by one time step , preparing to enter the calculation of the next simulation time step. This is the time advancement mechanism in simulation calculation. By continuously iterating this process, the dynamic behavior of the simulation system evolves over time.

[0214] Repeat steps 2 (monitoring process) to 6 (advancing simulation) until the simulation is complete: The monitoring process continues, along with trigger evaluation, target resolution determination, resolution switching (if necessary), and simulation advancement, until the simulation end condition is met. The simulation end condition can be set based on the specific simulation objectives and requirements, such as reaching a specified simulation time duration, the system reaching a stable operating state, or the occurrence of a specific event.

[0215] 8. Complete the simulation Saving results: After the simulation is complete, save the final state of the system and various data generated during the simulation, such as system state variables at different time points, event records, and resolution switching information. These saved results can be used for subsequent analysis and research, such as evaluating the system's dynamic performance, analyzing the impact of faults, and verifying the effectiveness of control strategies.

[0216] In step S400 of some embodiments, the target model resolution is assigned to the corresponding subsystem, and the subsystem of the simulation system is simulated and calculated based on the multi-level resolution model library and the adaptive resolution control strategy to obtain simulation calculation results corresponding to the power system containing high penetration renewable energy.

[0217] Specifically, step S400 may include but is not limited to the following steps: 1. Assign target model resolution to corresponding subsystems: 1. Analyzing Subsystem Characteristics: For power systems with high penetration rates of renewable energy, their subsystems exhibit distinct characteristics. For example, the output power of renewable energy generation subsystems (such as wind power and photovoltaics) is intermittent and fluctuating, and their dynamic characteristics are relatively rapid, potentially requiring higher resolution to accurately simulate these changes. Conventional power subsystems, such as traditional thermal power units, exhibit relatively slow dynamic response, necessitating a lower resolution under certain conditions.

[0218] The power load subsystem also has different characteristics depending on the load type (e.g., industrial load, commercial load, residential load, etc.). Generally speaking, large industrial loads vary relatively smoothly and can be assigned a relatively low resolution; while commercial and residential loads, due to their greater randomness and volatility, may require a higher resolution.

[0219] 2. Refer to the multi-resolution model library: Select the appropriate resolution level for each subsystem based on a pre-built multi-resolution model library. The multi-resolution model library contains models of various resolutions for different subsystem types and operating conditions. For example, for a wind turbine system, the model library may include low-resolution models for simulating average output characteristics over long time scales (such as hours), as well as high-resolution models for simulating turbulence characteristics and instantaneous power fluctuations over short time scales (such as seconds or milliseconds).

[0220] By analyzing the system requirements and computing resources in the current simulation scenario, the optimal resolution model is assigned to each subsystem from the model library. For example, when studying the transient stability of the power system, a high-resolution model might be assigned to the wind subsystem to accurately capture its rapid power changes during transient processes; while when studying long-term energy balance issues, a lower-resolution model might be assigned to the wind subsystem to improve computational efficiency.

[0221] 2. Simulation calculation based on adaptive resolution control strategy: 1. Initialize the simulation environment: Before starting the simulation, you need to set initial conditions, including the initial states of each subsystem (such as voltage amplitude, phase angle, frequency, etc.), the system's network topology, load level, and renewable energy output. At the same time, load the corresponding model parameters and calculation methods based on the resolution assigned to the subsystem.

[0222] For example, for a high-resolution subsystem model, a finer time step and a more complex numerical calculation method may be required to ensure the accuracy of the calculation; while for a low-resolution subsystem model, a larger time step and a relatively simple calculation method can be used.

[0223] 2. Simulation calculation process: The simulation calculation is carried out step by step according to the set time step. In each time step, the corresponding calculation and update are carried out according to the current system operation status and the resolution requirements of each subsystem.

[0224] For power systems with high penetration of renewable energy, particular attention must be paid to variations in the power output of the renewable energy subsystem. Due to the intermittent and fluctuating nature of renewable energy, their power output can fluctuate significantly over a short period of time. The adaptive resolution control strategy automatically adjusts its resolution based on detected variations in renewable energy power. For example, when rapid variations in wind power are detected, the system automatically switches to a high-resolution model for accurate simulation of its dynamics. When wind power variations are relatively stable, the system maintains a lower-resolution model for improved computational efficiency.

[0225] At the same time, considering the interactions and couplings between different subsystems in the power system, global state updates and solutions are required within each time step. For example, by solving the power system's power flow equations and rotor swing equations, updated values ​​for state variables such as the voltage at each node and the speed and power angle of each generator are obtained.

[0226] 3. Monitoring and Adjustment: During the simulation process, the operating status of each subsystem and the accuracy of the simulation results are continuously monitored. If the simulation results of a subsystem show large errors or do not meet the preset accuracy requirements, the resolution of the subsystem is dynamically adjusted according to the adaptive resolution control strategy.

[0227] For example, if during the simulation process it is found that the power changes of a certain load subsystem show abnormal fluctuations, resulting in local voltage instability, it may be necessary to increase the resolution of the load subsystem to more accurately simulate its load characteristics and voltage response; conversely, if the operating state of a certain subsystem is relatively stable and the accuracy of its simulation results has met the requirements, its resolution can be appropriately reduced to save computing resources.

[0228] 3. Get the simulation results: 1. Result summary and analysis: After the simulation is completed, the simulation calculation results of each subsystem are summarized, including the state variables at each moment (such as voltage, current, power, frequency, etc.), event records (such as fault occurrence time, recovery time, switching operation, etc.), resolution switching information, etc.

[0229] These results are analyzed in detail to assess various aspects of the performance of power systems with high renewable energy penetration. For example, voltage and frequency variations are analyzed to assess the system's power quality and stability; renewable energy utilization and curtailment are analyzed to assess the effectiveness of renewable energy absorption; and the interaction and coordinated operation of different subsystems are analyzed to assess the system's operational efficiency and reliability.

[0230] 2. Result Verification and Application: Compare the simulation results with the actual system operating data or other reliable reference data to verify the accuracy and effectiveness of the simulation model and calculation method. If a significant deviation is found between the simulation results and the actual situation, the simulation model and parameters need to be adjusted and optimized, and the simulation calculation needs to be repeated.

[0231] Verified simulation results can be applied in a variety of areas, such as power system planning and design, operation and dispatch, and control strategy formulation. For example, simulation results can be used to evaluate the impact of different renewable energy integration schemes on the power system, providing a basis for formulating reasonable energy development plans; or to optimize the power system's operation and dispatch strategies based on simulation results to improve renewable energy utilization and system operational efficiency.

[0232] In some embodiments, the physics-informed adaptive multi-resolution simulation framework provided by the present invention dynamically manages a model library across multiple resolution levels through an adaptive resolution control strategy (ARCS). It employs a multi-criteria triggering mechanism to perform seamless resolution switching with state mapping, and accurately identifies the perturbation-affected region through a hybrid influence domain identification strategy based on dynamic multi-metric evaluation. The integration of physics-informed neural networks as a unique resolution level increases flexibility and leverages advances in data-driven modeling.

[0233] Compared to full EMT, the expected benefits include a significant reduction in computational cost while maintaining high fidelity for key dynamic events, surpassing the capabilities of fixed-resolution hybrid methods. The proposed method significantly improves the computational efficiency of large-scale renewable energy power system simulations while maintaining the required simulation accuracy in critical areas.

[0234] In some embodiments of the present invention, a series of test experiments were designed to validate the effectiveness of the proposed method. The test systems included a modified IEEE 39-bus system and larger synthetic / real systems (>1000 nodes), equipped with detailed renewable energy / HVDC models, and developed PINN dynamic equivalent models for key subsystems.

[0235] The simulation scenarios include various faults, large disturbances, renewable energy control interactions, oscillations, and static periods to fully test the adaptability of the system. The performance evaluation indicators mainly include the following three aspects: 1. Accuracy: Compare to a full EMT (where feasible) or a highly detailed reference simulation. Use error metrics (e.g., root mean square error (RMSE), normalized RMSE, maximum absolute error (MAE)) to quantitatively measure the accuracy of key variables (key bus voltages, frequencies, line flows, renewable energy output, and control signals) for the duration of the simulation or during a specific dynamic event. Expected Results: The proposed method should demonstrate significantly higher accuracy than a fixed low-resolution (e.g., full RMS) simulation and comparable accuracy to a fixed hybrid simulation in key regions during the event, while closely matching the phenomena of a full EMT for high-resolution targets.

[0236] 2. Computational Efficiency: Measured by simulation runtime (CPU time, wall clock time) and potential memory usage. Speedup factors are calculated relative to full EMT and fixed-hybrid simulations. Expected Results: Significant speedup compared to full EMT (possibly orders of magnitude for large systems). Measurable speedup compared to fixed-hybrid simulations is achievable, especially when high-detail regions are unnecessarily large or remain static for extended periods.

[0237] 3. Adaptive Effectiveness: ARCS behavioral analysis: frequency and type of resolution switching, trigger activation statistics, and effectiveness of impact region identification. Visualization of dynamic resolution maps over time. Expected Results: Demonstrate that ARCS correctly identifies key events / dynamics, applies appropriate resolution increases in relevant regions, and reverts to lower resolution during periods of inactivity, thereby achieving the observed accuracy / efficiency balance. Comparison methods for this embodiment include: full EMT simulation (accuracy gold standard, accelerated baseline); full RMS simulation (low-accuracy, high-speed baseline); and fixed hybrid EMT-RMS simulation with intelligently selected but static partitions.

[0238] This comparison will highlight the benefits that derive specifically from the adaptability of the method of the present invention.

[0239] In a specific test case, a scenario in which a regional distribution network has a 50% renewable energy penetration rate was used as an example to simulate the system recovery process after a three-phase short-circuit fault. In the area near the fault point, the proposed method automatically upgrades to the EMT level simulation; a detailed RMS model is used in the affected area near the fault; and a simplified RMS or PINN dynamic equivalent model is used in more distant areas. As the fault is cleared and the system recovers, the dynamic resolution area gradually shrinks, and computing resources are concentrated in areas requiring high-precision simulation. The results show that compared with full EMT simulation, the proposed method achieves an approximately 20-fold acceleration in computing time, while the error in the voltage and current waveforms of key nodes does not exceed 2%, demonstrating the method's ability to significantly improve computing efficiency while maintaining key dynamic accuracy.

[0240] The present invention provides a method and device for simulating a power system based on adaptive multi-resolution physical information. The method and device obtain model data and operating status data of the power system, wherein the power system is a power system containing high-penetration renewable energy; construct a simulation system corresponding to the power system based on the model data, the operating status data, and a multi-level resolution model library, wherein the simulation system includes at least a plurality of different subsystems; determine the target model resolution corresponding to each subsystem through an adaptive resolution control strategy and the multi-level resolution model library; assign the target model resolution to the corresponding subsystem, and perform simulation calculations on the subsystems of the simulation system based on the multi-level resolution model library and the adaptive resolution control strategy to obtain simulation calculation results corresponding to the power system containing high-penetration renewable energy. The method is used to solve the contradiction between computational efficiency and accuracy in the simulation of large-scale power systems containing high-penetration renewable energy in the prior art. By designing an adaptive resolution control strategy, dynamic adjustment of the model resolution of different regions of the power system is achieved, which can significantly improve computational efficiency while maintaining simulation accuracy.

[0241] The following describes a physical information-based adaptive multi-resolution power system simulation device provided by the present invention. The physical information-based adaptive multi-resolution power system simulation device described below and the physical information-based adaptive multi-resolution power system simulation method described above can refer to each other.

[0242] like Figure 7 FIG. 1 is a schematic diagram of the structure of a physical information-based adaptive multi-resolution power system simulation device provided by the present invention. The physical information-based adaptive multi-resolution power system simulation device includes the following modules: An acquisition module 710 is used to acquire model data and operating status data of the power system; A construction module 720 is configured to construct a simulation system corresponding to the power system based on the model data, the operating status data, and a multi-resolution model library, wherein the simulation system includes at least a plurality of different subsystems; A determination module 730 is configured to determine a target model resolution corresponding to each subsystem using an adaptive resolution control strategy and the multi-level resolution model library; The simulation calculation module 740 is used to assign the target model resolution to the corresponding subsystem, and perform simulation calculation on the subsystem of the simulation system based on the multi-level resolution model library and the adaptive resolution control strategy to obtain simulation calculation results corresponding to the power system.

[0243] The embodiment of the physical information-based adaptive multi-resolution power system simulation device provided by the present invention is compatible with the embodiment of the physical information-based adaptive multi-resolution power system simulation method described above, and will not be described in detail here.

[0244] The present invention provides a method and device for simulating a power system based on adaptive multi-resolution physical information. The method and device obtain the model data and operating status data of the power system; construct a simulation system corresponding to the power system based on the model data, the operating status data and the multi-level resolution model library, wherein the simulation system includes at least a plurality of different subsystems; determine the target model resolution corresponding to each subsystem through an adaptive resolution control strategy and the multi-level resolution model library; assign the target model resolution to the corresponding subsystem, and perform simulation calculations on the subsystems of the simulation system based on the multi-level resolution model library and the adaptive resolution control strategy to obtain simulation calculation results corresponding to the power system. The method is used to solve the contradiction between computational efficiency and accuracy in the simulation of large-scale power systems containing high-penetration renewable energy in the prior art. By designing an adaptive resolution control strategy, dynamic adjustment of the model resolution of different regions of the power system is achieved, which can significantly improve computational efficiency while maintaining simulation accuracy.

[0245] Figure 8 An example of a physical structure diagram of an electronic device is shown below. Figure 8 As shown, the electronic device may include: a processor 810, a communications interface 820, a memory 830, and a communications bus 840, wherein the processor 810, the communications interface 820, and the memory 830 communicate with each other via the communications bus 840. The processor 810 may call logic instructions in the memory 830 to execute a method for adaptive multi-resolution power system simulation based on physical information, the method comprising: obtaining model data and operating status data of the power system; constructing a simulation system corresponding to the power system based on the model data, the operating status data, and a multi-level resolution model library, wherein the simulation system includes at least a plurality of different subsystems; determining a target model resolution corresponding to each subsystem through an adaptive resolution control strategy and the multi-level resolution model library; allocating the target model resolution to the corresponding subsystem, and performing simulation calculations on the subsystems of the simulation system based on the multi-level resolution model library and the adaptive resolution control strategy to obtain simulation calculation results corresponding to the power system.

[0246] Furthermore, the logic instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product, stored in a storage medium, includes instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0247] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the physical information-based adaptive multi-resolution power system simulation method provided by the above methods, the method including: obtaining model data and operating status data of the power system; constructing a simulation system corresponding to the power system based on the model data, the operating status data and a multi-level resolution model library, the simulation system including at least a plurality of different subsystems; determining the target model resolution corresponding to each subsystem through an adaptive resolution control strategy and the multi-level resolution model library; allocating the target model resolution to the corresponding subsystem, and performing simulation calculations on the subsystems of the simulation system based on the multi-level resolution model library and the adaptive resolution control strategy to obtain simulation calculation results corresponding to the power system.

[0248] On the other hand, the present invention also provides a non-transient computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the physical information-based adaptive multi-resolution power system simulation method provided by the above-mentioned methods, the method comprising: obtaining model data and operating status data of the power system; constructing a simulation system corresponding to the power system based on the model data, the operating status data and a multi-level resolution model library, the simulation system comprising at least a plurality of different subsystems; determining the target model resolution corresponding to each subsystem through an adaptive resolution control strategy and the multi-level resolution model library; allocating the target model resolution to the corresponding subsystem, and performing simulation calculations on the subsystems of the simulation system based on the multi-level resolution model library and the adaptive resolution control strategy to obtain simulation calculation results corresponding to the power system.

[0249] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0250] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0251] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A physical information-based adaptive multi-resolution power system simulation method, characterized in that: include: Obtain model data and operating status data of the power system; Building a simulation system corresponding to the power system based on the model data, the operating status data and a multi-resolution model library, wherein the simulation system includes at least a plurality of different subsystems; Determining the target model resolution corresponding to each subsystem through the adaptive resolution control strategy and the multi-level resolution model library; The target model resolution is allocated to the corresponding subsystem, and simulation calculations are performed on the subsystems of the simulation system based on a multi-level resolution model library and an adaptive resolution control strategy to obtain simulation calculation results corresponding to the power system.

2. The method for adaptive multi-resolution power system simulation based on physical information according to claim 1, characterized in that: The multi-level resolution model library includes at least an electromagnetic transient model, a detailed root mean square model, a simplified root mean square model, a physical information neural network dynamic equivalent model, a quasi-steady-state model, and a fixed load model; The resolution level definition of the adaptive resolution control strategy includes at least: a first accuracy level corresponding to the electromagnetic transient model, a second accuracy level corresponding to the detailed root mean square model, a third accuracy level corresponding to the simplified root mean square model, a fourth accuracy level corresponding to the physical information neural network dynamic equivalent model, and a fifth accuracy level corresponding to the quasi-steady-state model and / or the fixed load model; The first level of accuracy is characterized by the use of an electromagnetic transient model with a time step of 1-50s to capture switching transients, harmonics, and fast control dynamics; The second accuracy level is characterized by the use of a detailed RMS model with a time step of 1-5 ms to capture electromechanical oscillations and detailed renewable energy control dynamics; The third level of accuracy is characterized by using a simplified RMS model with a time step of 5-10ms to capture frequency dynamics and initial voltage stability; The fourth level of accuracy is characterized by the use of a physical information neural network dynamic equivalent model with a time step greater than 1ms, which is used to represent aggregate dynamic behavior and learn complex interactions; The fifth accuracy level is characterized by using a quasi-steady-state model and / or a fixed load model, a time step greater than 100 ms, and only minimal dynamic characteristics.

3. The method for adaptive multi-resolution power system simulation based on physical information according to claim 2, characterized in that: Determining the target model resolution corresponding to each subsystem through the adaptive resolution control strategy and the multi-level resolution model library includes: Based on the multi-criteria triggering mechanism in the adaptive resolution control strategy and the multi-level resolution model library, the target model resolution corresponding to each subsystem is determined; wherein, the target model resolution is one of the first accuracy level, the second accuracy level, the third accuracy level, the fourth accuracy level and the fifth accuracy level determined based on the multi-criteria triggering mechanism.

4. The method for adaptive multi-resolution power system simulation based on physical information according to claim 3, characterized in that: The multi-criteria triggering mechanism in the adaptive resolution control strategy includes: An event-based trigger, wherein the event-based trigger is used to respond to discrete events detected by the monitoring module of the simulation system, wherein the discrete events include at least faults, topology changes, control mode switching, and protection actions; A state-based trigger, wherein the state-based trigger is used to simulate system variables by continuously monitoring the system, wherein the system variables include at least voltage deviation, frequency deviation, power oscillation detection and key control signal activity; a sensitivity-based trigger for evaluating a potential impact on a region of interest, wherein the potential impact includes at least an impact on the region of interest and a model difference metric; A stillness detection trigger is used to detect minimum dynamic activity of a high-resolution subsystem within a duration when the resolution is reduced.

5. The method for adaptive multi-resolution power system simulation based on physical information according to claim 4, characterized in that: After the step of assigning the target model resolution to the corresponding subsystem, the method includes: Perform smooth resolution switching with state variable mapping to achieve seamless transitions between multi-resolution models; The step of performing smooth resolution switching with state variable mapping comprises: In the conversion between the root mean square model and the electromagnetic transient model, simulation continuity is ensured by mapping between phasors and instantaneous values; wherein the root mean square model is one of the detailed root mean square model and the simplified root mean square model; In the conversion between different levels of RMS models, the common state is transferred and the remaining state variables are processed; In the conversion between the root mean square model and the dynamic equivalent model of the physical information neural network, the state is initialized based on the input relationship or output relationship of the neural network; Use smooth transition or projection methods to handle discontinuities in smooth resolution switching and minimize numerical transients.

6. The method for adaptive multi-resolution power system simulation based on physical information according to claim 1, characterized in that: The allocating the target model resolution to the corresponding subsystem includes: Dynamically allocate target model resolution to different subsystems based on real-time system events, state deviations, and sensitivity analysis; The dynamic allocation of target model resolutions to different subsystems based on real-time system events, state deviations, and sensitivity analysis also includes: A hybrid impact domain identification strategy based on multi-index dynamic evaluation is used to determine the disturbance impact area; The hybrid impact domain identification strategy includes: Initial rapid delineation stage: combining topological information with simplified electrical distances to quickly determine the initial candidate impact domain within a very short time after the disturbance occurs; Dynamic boundary monitoring and adjustment phase: Dynamically expand or shrink the impact domain by real-time monitoring of multiple indicators such as the state deviation gradient, power flow change rate, local frequency change rate, and control signal activity of boundary nodes and neighboring nodes; Stabilization and regional management stage: When the main dynamic process of the disturbance tends to ease, the impact domain is finally stabilized based on the mechanism of impact domain stability confirmation, impact domain merging and final shrinkage / elimination of the impact domain.

7. The method for adaptive multi-resolution power system simulation based on physical information according to claim 6, characterized in that: The process of shrinking the influence domain in the dynamic boundary monitoring and adjustment stage includes: Define the core impact area: Identify the area around the initial incident source or the area containing critical equipment as the core area that will not participate in the contraction; Identify candidate shrinkage areas: Identify components / subsystems that are inside the candidate influence domain but outside the core influence area as candidate shrinkage areas; Evaluating the dynamics of the candidate contraction region: monitoring local static indicators of the components within the candidate contraction region, wherein the local static indicators include at least: whether the rate of change of the state variable is lower than a threshold, whether the deviation from the quasi-steady-state value is lower than a threshold, and whether the relevant control signal is inactive; Evaluate the connectivity of candidate shrinkage regions: Check whether removing a candidate shrinkage region would fragment the remaining high-resolution regions or sever the connection between the core influence region and the current active boundary; Perform shrinkage: When the candidate shrinkage area meets the local static conditions, does not belong to the core influence area, and its removal does not destroy the necessary connectivity, the corresponding resolution is reduced and removed from the current high-resolution influence domain.

8. The method for adaptive multi-resolution power system simulation based on physical information according to claim 1, characterized in that: The steps of constructing the physical information neural network dynamic equivalent model include: Offline training is performed using high-precision simulation data, and physical constraints are incorporated into the loss function during training. Predict the subsystem's response based on boundary conditions online to provide a dynamic representation of the simulated system; The switching between the physical information neural network dynamic equivalent model and the remaining resolution level models is achieved through an adaptive resolution control strategy.

9. The method for adaptive multi-resolution power system simulation based on physical information according to claim 6, characterized in that: The steps of the simulation calculation include: Initialization processing: setting the initial time, initial state and initial resolution; Monitoring and processing: obtaining current status and detecting events; Evaluation triggers: Identify subsystems that require resolution changes based on a multi-criteria trigger mechanism; Determine the target resolution: For each subsystem that needs to change the resolution, determine the new resolution level; Switch resolution: Pause the subsystem whose resolution needs to be switched, obtain the current state, load the new multi-resolution model, map the state variables, update the interface and resume the simulation; Advance simulation: Solve the simulation system equations using the current resolution configuration, obtain the solution, and update the state based on the solution; Repeat the above monitoring process to advance the simulation steps until the simulation ends.

10. A physical information-based adaptive multi-resolution power system simulation device, characterized in that: include: An acquisition module, used to acquire model data and operating status data of the power system; a construction module, configured to construct a simulation system corresponding to the power system based on the model data, the operating status data, and a multi-resolution model library, wherein the simulation system includes at least a plurality of different subsystems; A determination module, configured to determine a target model resolution corresponding to each subsystem by using an adaptive resolution control strategy and the multi-level resolution model library; The simulation calculation module is used to allocate the target model resolution to the corresponding subsystem, and perform simulation calculation on the subsystem of the simulation system based on the multi-level resolution model library and the adaptive resolution control strategy to obtain the simulation calculation results corresponding to the power system.

Images