A power system distributed simulation method based on data distribution service

Abstract

The application discloses a power system distributed simulation method based on a data distribution service, and belongs to the technical field of power system simulation modeling. The method comprises the following steps: decoupling a power system model in mathematics, dividing the power system model into interconnected subsystems, and deducing a unified mathematical interface between the subsystems; determining a data transmission scheme between the subsystems, designing and realizing a communication interface by using a data distribution service (DDS); determining a synchronization scheme between different time scale subsystems; and realizing data transmission and simulation between multiple devices. Finally, a distributed joint simulation platform based on the DDS is constructed. The application overcomes the contradiction between simulation accuracy and simulation efficiency in a traditional centralized simulation mode, and can meet the multi-dimensional requirements of a new type of power system analysis on fine model accuracy, efficient simulation rate and general simulation architecture expansibility.

Description

Technical Field

[0001] This invention belongs to the field of power system modeling and simulation technology, and more specifically, relates to a distributed simulation method for power systems based on data distribution services. Background Technology

[0002] With the continuous grid integration of renewable energy and the widespread application of power electronic equipment, modern power systems are transforming into new power systems with a high proportion of new energy integration, resulting in increasingly complex structures. Their operation, control, and analysis heavily rely on power system simulation technology. However, the integration of new energy and power electronic equipment places higher demands on the refinement of simulation modeling. Furthermore, the large-scale spatiotemporal coupling characteristics of new power systems reduce the simulation timescale and complicate dynamic processes, leading to a significant increase in simulation time. Simultaneously, limited by the availability of computational resources, existing simulation technologies struggle to balance rapid simulation computation with refined modeling, highlighting the growing contradiction between simulation accuracy and efficiency. In addition, new power systems face practical demands for larger simulation scales, smaller simulation step sizes, and more simulation scenarios. Therefore, existing simulation technologies have certain limitations.

[0003] Currently, power system simulations mostly adopt a centralized approach, meaning that modeling and analysis are carried out using a single simulation software. Some studies reduce model complexity and shorten simulation time by optimizing and simplifying the model, but this is difficult to meet the accuracy requirements of operation and control. Other studies utilize parallel computing technology, but these studies are still essentially centralized solutions and still have several problems: first, the computational load is large, relying on supercomputers; second, idle computing resources are not fully utilized; and third, they are limited to a single software platform, failing to meet the diverse simulation needs and performance optimization requirements.

[0004] Distributed simulation technology offers a new approach to power system simulation. It boasts the advantage of parallel computing for improved efficiency and uses task combination as its underlying logic, eliminating the need for prior knowledge of the entire large system. Furthermore, it allows for flexible integration of subsystems after the architecture is built. In the field of power system simulation, distributed architecture is primarily used for solving differential-algebraic equations. It enables refined modeling and improves simulation speed in terms of computational efficiency. However, existing simulation technologies still suffer from several problems: (1) they do not adequately consider the scalability of modeling and simulating new power systems; (2) the interface design between subsystems and the data transmission process are complex; and (3) the simulation computation speed remains relatively slow. Therefore, how to construct a highly compatible and scalable distributed simulation architecture is a pressing technical issue that needs to be addressed. Summary of the Invention

[0005] In view of the above-mentioned limitations or improvement needs of the existing technology, the present invention provides a distributed simulation method for power systems based on data distribution services, which aims to improve the scalability and simulation efficiency of large-scale new power system simulation.

[0006] To achieve the above objectives, the present invention provides a distributed simulation method for power systems based on data distribution services, comprising:

[0007] The power system is divided into n+1 subsystems, including n generator subsystems and 1 network subsystem. A data distribution service is used to design communication interfaces to realize data exchange between subsystems. The nodes of the n+1 subsystems are mapped to at most n+1 devices. The subsystems are jointly simulated through the devices, and the devices can interact with each other.

[0008] The generator subsystems will perform parallel computations, while the generator subsystem and the network subsystem will solve problems alternately.

[0009] Furthermore, the joint simulation of the subsystem via the device includes:

[0010] Get the start signal;

[0011] Each subsystem receives data;

[0012] According to the preset simulation time step, a step signal is sent to control the timing of subsystem simulation calculation and data exchange until all subsystems complete the preset simulation time.

[0013] Each subsystem sends data;

[0014] Obtain the stop signal.

[0015] Furthermore, this invention overcomes the limitations of traditional centralized modeling, which requires a unified system-wide architecture. Based on the decoupling requirements between the physical characteristics of the power system and the mathematical model, and utilizing the matrix storage capabilities of Matlab / Simulink, the complete system is divided into n generator subsystems and one network subsystem. Each subsystem can be independently built and solved on different devices. To ensure that the decomposed model maintains both local modeling accuracy and the ability to transmit key parameters through interfaces, the generator subsystem uniformly uses generator output current and equivalent equipment admittance for transmission via interfaces, while the network subsystem uniformly uses network node voltages for transmission via interfaces, thus restoring the physical coupling relationships of the entire system.

[0016] In a distributed simulation architecture, subsystem partitioning should aim to balance the simulation time consumption of each device across subsystems, thereby shortening the total parallel computing time and maximizing the efficiency of distributed simulation. This paper proposes a subsystem partitioning scheme based on time consumption balancing, which balances the simulation time consumption of each subsystem in a single round, achieving optimal matching of computing resources.

[0017] First, based on differences in system physical topology and dynamic characteristics, the system is initially divided into... There are three basic subsystems. Let the number of node devices that can participate in the simulation calculation be . When satisfied For example, the number of subsystems that can typically be divided into for a small-scale system. If the number of subsystems is small, then one device is allocated to each subsystem, and the subsystems are simulated in parallel. At that time, it is necessary to The subsystems are integrated into Group. If the total simulation time of the subsystems on each group's equipment... Significant differences can lead to situations where "fast subsystems wait for slow subsystems," resulting in some devices experiencing substantial idle time. Therefore, it's necessary to adjust the subsystem partitioning method to achieve this. The overall simulation time of the subsystems on the group of devices tends to be balanced, maximizing the efficiency of parallel computing. For each subsystem, the runtime of a single simulation round is defined. as follows:

[0018]

[0019] in, For single-round simulation of neutron system The number of simulation steps performed. For subsystem The single-step simulation time, This refers to the time for a single API call. Let be the duration of a single communication. Let the be... The set of subsystems included in the group device is The subsystem partitioning objectives are defined as follows:

[0020]

[0021] First try to The basic subsystems are combined as follows Group, make Total simulation time on the group of devices Similar; if the groups are similar If the gap remains significant, further adjustments to the subsystem partitioning scheme are needed, including integrating some subsystems and recalculating communication latency and other parameters. and Through multiple rounds of iteration, the time consumption difference between each group was gradually reduced, ultimately making... Total simulation time on the group of devices Approaching equilibrium. It should be noted that the subsystem partitioning of a power system must take into account the coupling relationships between systems, and cannot be solved entirely as an unconstrained optimization problem. Therefore, when the idle waiting time of each device's parallel computation is minimized through the above adjustments, the overall parallel simulation efficiency of the system can be considered to have reached its optimal state.

[0022] Generator Subsystem: Separate differential equation models are built for synchronous generators, wind turbines, and other equipment (considering transient processes of the excitation and speed control systems), supporting independent local debugging and parameter optimization. The state equation is expressed as:

[0023]

[0024] in, and Representing subsystems The state variables and algebraic variables of dynamic behavior, 1≤i≤n.

[0025] Network Subsystem: This subsystem models the algebraic equations of an AC power grid (such as nodal voltage equations), focusing on power flow calculations and topology change response. The network equation is expressed as:

[0026]

[0027] Furthermore, this invention breaks away from the drawback of the traditional centralized simulation system's mandatory uniform minimum step size, allowing each simulation subsystem model to select its own step size based on its dynamic characteristics: for fast dynamic generator subsystems containing new energy inverters and energy storage converters, a small step size is used to accurately capture high-frequency transient processes; for slow dynamic subsystems of traditional synchronous generators, a large step size is used to reduce redundant calculations; for network subsystems, the step size is matched to that of the slowest generator subsystem to avoid efficiency loss due to frequent synchronization.

[0028] Furthermore, during the initialization phase, a start signal is sent to each subsystem to ensure that all subsystems start the simulation synchronously. During the running phase, the simulation progress of each subsystem is controlled, not using actual time for synchronization, but using simulation time steps as the unit. According to the preset simulation steps, signals are sent to schedule the simulation progress of the subsystems, and each subsystem performs step-by-step simulation within each simulation time step. During the termination phase, after confirming that all subsystems have completed the simulation for the specified duration, a simulation stop signal is sent to each subsystem.

[0029] Furthermore, after receiving the step simulation signal, each simulation subsystem does not start the simulation calculation directly. Instead, it ensures that each subsystem obtains all the necessary input data at the current simulation moment before starting the calculation. Otherwise, the process will automatically block and wait. When the missing data is received, the subsystem simulation resumes normal operation.

[0030] Furthermore, a coarse-grained parallel computing mode based on CPU is adopted among the generator subsystems: multiple devices can simultaneously support multiple generator subsystems, each independently solving the corresponding differential-algebraic equation system, significantly reducing the overall time consumption of a single time step through parallel computing. At the data storage and synchronization level, the core data of each simulation subsystem (such as state variables and intermediate calculation results) are stored locally. Only after each simulation step is completed, the interaction parameters (such as generator output current and instantaneous node voltage values) are synchronized in real time to the associated subsystems of other devices through coupled communication interfaces. This reduces redundant data storage across devices and ensures the accuracy of the coupling relationship between subsystems.

[0031] Furthermore, to ensure the privacy and security of information in each subsystem, the data that needs to be exchanged between subsystems is categorized according to its physical meaning and registered as DDS data topics. Each subsystem subscribes to the topics it needs to send and receive, ensuring the effective flow of simulation data while enabling selective data interaction between subsystems. This is particularly suitable for distributed simulation scenarios of power systems involving multiple stakeholders.

[0032] The present invention also provides an electronic device, comprising: a computer-readable storage medium and a processor;

[0033] The computer-readable storage medium is used to store executable instructions;

[0034] The processor is used to read executable instructions stored in the computer-readable storage medium and execute the above-described method.

[0035] The present invention also provides a computer-readable storage medium storing computer instructions for causing a processor to perform the above-described method.

[0036] The present invention also provides a computer program product, including a computer program or instructions that, when executed by a processor, implement the above-described method.

[0037] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:

[0038] 1. This invention transforms the traditional centralized simulation model into a distributed architecture simulation model. Based on the physical characteristics and mathematical decoupling principle, the complete system is divided into a generator subsystem and a network subsystem that can be solved independently. Since necessary data exchange is still required between the various subsystems after division, a unified interface between the subsystems is designed, and the key parameters and their transmission forms are clarified. This invention designs a top-level scheduling and subsystem communication interface, and realizes distributed data interaction by integrating data reception synchronization and data transmission.

[0039] 2. Compared to mainstream centralized simulation technologies, this invention allows each subsystem to autonomously select the optimal time step, avoiding waste of computing resources. Simultaneously, the parallel simulation among the distributed generator subsystems accelerates the simulation rate. Compared to existing distributed simulation technologies, this invention maintains high accuracy under different subsystem partitioning methods, imposing fewer constraints on simulation modeling. The synchronization method of top-level scheduling and internal subsystem synchronization achieves efficient collaboration in distributed simulation without timestamp alignment. Furthermore, the communication interface design between various subsystems is unified, facilitating integration and providing excellent scalability. Attached Figure Description

[0040] Figure 1 The overall architecture diagram of a novel distributed simulation of a power system based on data distribution services is provided for the example of this invention.

[0041] Figure 2 A diagram illustrating the alignment and synchronization scheme of the distributed simulation architecture provided as an example of the present invention.

[0042] Figure 3 A flowchart illustrating the solution of the differential algebraic equations system of the subsystem in the distributed simulation architecture provided as an example of the present invention.

[0043] Figure 4 The communication interface design diagram of the distributed simulation architecture provided for the example of this invention.

[0044] Figure 5 This diagram illustrates a distributed architecture implementation scheme for communication between multiple devices, as provided in this invention example.

[0045] Figure 6 The topology diagram of the IEEE New England 10-machine 39-node test system provided as an example of the present invention.

[0046] Figure 7 The topology diagram of the WECC 29-machine 179-node test system provided for the example of this invention.

[0047] Figure 8 Error graph of simulation results of the distributed architecture provided in this invention for the New England 10-machine 39-node test system.

[0048] Figure 9 A comparison chart of the overall execution time of the distributed architecture provided in this invention with that of centralized simulation at different scales. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0050] The following describes the distributed simulation architecture of power systems based on data distribution services provided in the embodiments of this application.

[0051] Figure 1 This is a schematic diagram of a novel distributed simulation architecture for a power system based on a data distribution service, according to an embodiment of this application. Figure 1 As shown, this distributed architecture mathematically decouples traditional power system simulation models, dividing them into multiple subsystems that can be solved independently and in parallel. Each subsystem is modeled separately and deployed on different devices as independent processes for simulation, utilizing multiple computing resources to achieve parallel solving (e.g., ...). Figure 1 As shown in (a)). During the simulation, the divided generator subsystems will be computed in parallel, while the generator subsystem and the network subsystem need to be solved alternately. Based on this, the present invention designs a communication interface and key parameters that need to be transmitted (such as...). Figure 1 As shown in (b)); since necessary data exchange is still required between the various subsystems after the division, this invention designs a communication interface between the subsystems based on a data distribution service. Each subsystem transmits data through the communication interface to achieve mutual data exchange (e.g., Figure 1 (as shown in (c)).

[0052] This invention provides a distributed simulation method for power systems based on data distribution services, comprising:

[0053] The power system is divided into n+1 subsystems, including n generator subsystems and 1 network subsystem. A data distribution service is used to design communication interfaces to realize data exchange between subsystems. The nodes of the n+1 subsystems are mapped to at most n+1 devices. The subsystems are jointly simulated through the devices, and the devices can interact with each other.

[0054] The generator subsystems will perform parallel computations, while the generator subsystem and the network subsystem will solve problems alternately.

[0055] Specifically, transient stability simulation of large-scale power systems can be represented as the time-domain solution of thousands of nonlinear differential-algebraic equations. Neglecting the effects of electromagnetic transient processes in the generator stator and power grid, the system process can be represented by a set of differential-algebraic equations, the general form of which is:

[0056]

[0057] in, Represents the state variables of the system. The first row of the equation represents the algebraic variables of the system; the differential equations in the first row represent the dynamic characteristics of the power system equipment, which are the state equations of the system; the second row of the equation represents the static characteristics of the power system components, which are mainly the network equations of the system.

[0058] In traditional centralized simulations, the entire system is built uniformly, and the aforementioned differential-algebraic equations are solved uniformly. To meet the stringent requirements for solution speed and storage space when solving large-scale differential-algebraic equations in power systems, a parallel computing architecture can be introduced to improve computational efficiency. When using a parallel architecture for power system simulation, the power system first needs to be dynamically decoupled. Referring to the physical meaning of each equation, the vast and complex power system is precisely divided into multiple independent subsets. The equations to be solved in the power system simulation can then be correspondingly divided into the following formula: Group:

[0059]

[0060] Components within the divided subsystems exhibit strong interdependencies, while components between different subsystems have only weak coupling. To ensure that the decomposed model maintains local modeling accuracy, key parameters are passed between subsystems through interfaces to reconstruct the physical coupling relationships of the entire system.

[0061] Specifically, traditional centralized simulation methods typically select the minimum time step required by each component in the system as a uniform standard. This approach may lead to unnecessary computational burden for the slower-moving parts of the system. In contrast, distributed simulation architectures optimize the use of computational resources by decoupling the system into independent subsystems, allowing each subsystem to choose the most suitable time step based on its characteristics.

[0062] Specifically, a top-level scheduling module of a distributed architecture is constructed to achieve unified management of the simulation processes of subsystems distributed across different devices. This module performs synchronization control in units of simulation time steps. This scheduling mechanism does not rely on actual physical time, but rather on simulation logical time for synchronization, effectively avoiding problems caused by clock deviations between different devices. The specific workflow is as follows:

[0063] Initialization phase: Start signals are sent to each subsystem, and the subsystems start simulation synchronously.

[0064] Operation phase: Control the simulation progress of each subsystem and send step signals according to the preset simulation step size. In each time step, control the timing of subsystem simulation calculations and data exchange to ensure simulation synchronization of different subsystems.

[0065] Termination phase: After confirming that all subsystems have completed the simulation for the specified duration, a stop signal is sent to each subsystem.

[0066] Specifically, data receiving and synchronization modules are integrated into each subsystem model to ensure that the subsystem only starts calculation after acquiring all necessary input data for the current simulation moment; otherwise, it automatically blocks and waits. The specific implementation method is as follows:

[0067] Before the subsystem performs the simulation calculation for the current step, it checks the currently received data. First, it checks whether the received data is the simulation result of the previous time step (based on the simulation time), and verifies whether all the data required for the current time step has arrived. If the data is complete, the subsystem uses this data as input to start the simulation calculation for one step. If the data is not updated or there is missing data, the subsystem simulation process is blocked until updated data or missing data is received. At that time, the integrity is rechecked, and if the conditions are met, the blocking is immediately lifted, and the subsystem simulation will continue to calculate based on the latest data.

[0068] Specifically, for subsystems with different time lengths, combined with Figure 2 This section introduces a simulation synchronization scheme for a distributed architecture. Assume there are two subsystems in the distributed architecture with different simulation time steps. Subsystem 1 has a simulation time step of... The simulation time step for subsystem 2 is ,and After receiving the necessary data, both subsystems simultaneously begin a simulation: Subsystem 1 executes... Subsystem 1 performs one step of simulation, while subsystem 2 performs one step of simulation. Once both subsystems have completed their simulations, data is exchanged to ensure the synchronization of the entire system's simulations. In actual simulations, the subsystem that completes its simulation first will remain in this state, waiting.

[0069] Specifically, when a distributed architecture is applied to power system simulation, multiple devices simultaneously carry multiple subsystem simulation models, each independently solving a system of differential-algebraic equations, thus achieving efficient distributed simulation. The specific solution is as follows:

[0070] Each subsystem after the division is independent of the others. Each subsystem has its own set of differential-algebraic equations, which are solved independently. For each subsystem... Given the necessary initial values, its transient stability simulation process can be expressed as:

[0071]

[0072] in, and These are representative subsystems State variables and algebraic variables of dynamic behavior, i.e., subsystems The internal variables; and and These represent the state variables and algebraic variables from other subsystems, respectively, i.e., the subsystems. External variables.

[0073] At each time step, the trapezoidal integral method is applied to the ordinary differential equations of each subsystem. Using the Gauss-Jacobi iterative scheme, the numerical solutions for each time step are obtained as follows:

[0074]

[0075] When subsystem Start a new time step ( When solving for ), the external variables required from other subsystems were already obtained at the previous time step ( Ready. Therefore, within each time step, The subsystems can be solved in complete parallel, without needing to follow a specific computation order. Therefore, in both centralized and distributed solutions, since the initial conditions of the systems are consistent and the network algebraic equations described are essentially equivalent, the results obtained by the two solution methods can be considered consistent as long as the solution error is effectively controlled within an acceptable range at each simulation time step.

[0076] Combination Figure 3 The simulation calculation process of each time step of the subsystem can be divided into three stages: data reception, independent solution, and result update. First, in the data reception stage, each subsystem obtains the required input variables through an interface; these variables depend on the outputs of other subsystems in the previous time step. In the independent solution stage, each subsystem performs independent calculations using its own state variables and input vectors. Finally, in the result update stage, after completing the calculation, the subsystem updates its output variables through an interface and exchanges these instantaneous values ​​with other subsystems to ensure the coordinated operation of the system and the accuracy of the results.

[0077] More intuitively, for subsystems Subsystem The relationship between the input and the output of the remaining subsystems is represented as follows:

[0078]

[0079] in, It is a subsystem In time The input vector, Is subsystem i in time The output vector, This indicates the correspondence between subsystems.

[0080] From the perspective of solving differential equations, the data exchange relationship between the systems is as follows: First, the state variables and algebraic variables of the system are initially estimated and assigned initial values. Then, at each time step... All Each subsystem is solved independently. Once the iterative solutions of all subsystems reach convergence, the system updates the values ​​of these variables and exchanges state and algebraic variables between the subsystems.

[0081] In addition, regarding data storage, the simulation data of each subsystem (such as state variables, algebraic variables, and intermediate calculation results) is stored on local devices. Only after each simulation step is completed, the interactive parameters (such as generator output current and instantaneous node voltage values) are synchronized to the associated subsystems of other devices through the coupled communication interface.

[0082] Figure 4 This is a design diagram of the communication interface for the distributed simulation architecture provided in this invention. The core data of each simulation subsystem (such as state variables and intermediate calculation results) are stored on the local device. Only after each simulation step is completed, the interaction parameters (such as generator output current and instantaneous node voltage values) are synchronized in real time to the associated subsystems of other devices through the coupled communication interface. This reduces redundant data storage across devices and ensures the accuracy of the coupling relationship between subsystems.

[0083] Specifically, the data transmission scheme of the distributed architecture adopts a publish-subscribe architecture, such as... Figure 5 In a publish-subscribe architecture, the subsystems sending and receiving data are decoupled. There's no need for one-to-one communication connections between subsystems; instead, all connected simulation programs are treated as a virtual "data bus," allowing direct communication between subsystems without the need for intermediaries like servers. In power system simulation, the information between the divided subsystems doesn't need to be fully shared. For example, a generator subsystem doesn't need to receive simulation iteration results from all other generator subsystems. Therefore, the distributed architecture divides all data to be transmitted into several groups, corresponding to parameters like current and voltage. Each group is registered with a unique "data topic" within its domain, and publishers and subscribers are configured for each subsystem. Publishers upload data to the corresponding data topic; subscribers can subscribe to data topics of interest, thus receiving only the data information needed by that subsystem. Therefore, the distributed architecture only needs shared data topics for message transmission. This data transmission scheme significantly reduces system coupling and improves the flexibility and scalability of the distributed architecture.

[0084] Combination Figure 5This paper introduces a communication implementation scheme for simulating distributed architecture across multiple devices. When a new subsystem is added to the distributed architecture, it is not necessary to explicitly declare its existence to other subsystems; instead, its communication interface can be directly coupled and the corresponding data object configured. Specifically, the new subsystem is first assigned to the DDS domain of the current system to establish a communication foundation. As a participant within the domain, the new subsystem needs to couple with the DDS module. By creating the appropriate number of publishers and subscribers and selecting suitable data topics, it can complete data publishing and subscription to access the distributed communication network. If the new subsystem needs to transmit data through a new data topic, a new topic can be registered based on the data type to be transmitted, and configuration parameters can be updated for the new subsystem and the subsystems exchanging data with it. Through the above process, the new subsystem can efficiently communicate with other subsystems in the distributed architecture.

[0085] Ultimately, this invention, based on the MatPSST toolkit, completed the construction of a distributed simulation verification platform in Matlab / Simulink. The system simulation model was solved in Matlab / Simulink, which provides various ODE solvers to meet the needs of solving the system's differential-algebraic equations. In the development of the distributed architecture communication interface, this invention used RTI DDS version 6.1.1 to complete the data topic registration, communication entity (publisher and subscriber) creation, and data publish-subscribe mechanism. Furthermore, it programmed the coupling between the DDS communication module and the subsystem simulation model, enabling the simulation model to call DDS through a unified communication interface to achieve data transmission between subsystems.

[0086] To evaluate the accuracy of the proposed distributed simulation architecture, this invention uses the average error as a benchmark against the results of traditional centralized simulations. As a performance metric, it is used to quantitatively evaluate the overall relative error between distributed and centralized simulation results over a period of time. For simulation... m For simulation data curves at a given time step and They represent the first Centralized simulation results and distributed simulation results at each time step. This represents the average value of the centralized simulation results over the entire period. The closer the value is to 0, the more accurate the distributed simulation results are.

[0087]

[0088] To scientifically and intuitively evaluate the performance of the proposed distributed simulation architecture, the overall simulation execution time is selected as the evaluation metric for quantitative assessment. The overall simulation execution time is the actual time taken from simulation initialization to the end of the entire system simulation. For traditional centralized simulations, only the actual time consumed by a single simulation software needs to be calculated. For distributed simulation, define its actual time consumption. It mainly consists of three parts: the parallel simulation time of all subsystems. Communication time between subsystems and API call time By comparing the overall execution time of the two simulation modes in the same scenario, the time efficiency of the distributed architecture can be intuitively evaluated. When the time saved by parallel simulation of the distributed architecture exceeds the additional time added by system communication and waiting, that is... In this context, distributed architecture offers advantages in simulation efficiency. Furthermore, in distributed simulations, the process execution time of each subsystem's host computer is used as a quantitative indicator of computational resource consumption. process runtime definition This includes the simulation, communication, and interface call times of the subsystem. The simulation time here only considers the simulation time of this specific subsystem, not the total simulation time of all parallel subsystems. Subsystem process runtime quantifies the resource occupancy of the device—when computing resources are idle, they can be allocated to other tasks, achieving optimal resource allocation. Furthermore, this metric reflects the rationality of the subsystem partitioning—when the runtime of subsystems on different devices tends to be balanced, it indicates that the current subsystem partitioning method achieves optimal parallel computing performance. Considering the differences in real-time computer operating states and fluctuations in network communication conditions, which may lead to variations in the time taken for a single measurement, the average of multiple simulation results was recorded in subsequent simulation verifications.

[0089] The above method will be explained in detail below using the IEEE New England 10-machine 39-node test system and the WECC 29-machine 179-node test system as examples. The system topology is as follows: Figure 6 and Figure 7 .

[0090] For the WECC29 179-node test system, simulation examples with different numbers of subsystems and different decoupling methods were set during the system partitioning phase. The system was divided into 2 and 4 generator subsystems according to the generator node number, respectively. Different system fault conditions were set during the simulation test, with a three-phase short circuit fault occurring at bus 7 or bus 101 at 1 second. The simulation test results show that the location of the fault has a certain impact on the simulation results. Changing the fault conditions causes the average error of the simulation results to fluctuate slightly within 5%, but it remains within an acceptable range. Under the same fault conditions, the simulation error fluctuation caused by increasing the number of subsystems is small. When the number of subsystems increases from 2 to 4, the increase in average error does not exceed 0.1%, which can be considered that the accuracy of the proposed distributed architecture is almost unaffected by different decoupling and system partitioning methods. The above results show that the distributed architecture has wide applicability and reliability in practical applications. The number of subsystems can be expanded or reduced according to actual conditions, effectively adapting to different system compositions and maintaining high accuracy.

[0091] Furthermore, this invention compares the test results of the IEEE four-machine two-zone system, the improved New England 10-machine 39-node system, and the WECC 29-machine 179-node system from two dimensions: simulation accuracy and simulation efficiency. Regarding simulation accuracy, Figure 8 compares the overall average error distribution of the New England 10-machine 39-node system and the WECC 29-machine system. The average error of the former is concentrated at 0.1%, while the average error of the latter is concentrated at around 1%. As the system scale increases, the simulation error shows a slight increase, but it remains within an acceptable error range.

[0092] In terms of simulation efficiency, the improvement in simulation speed compared to centralized simulation was compared when distributed simulation architecture was applied to systems of different scales. For the comparison, simulations of systems of different scales were selected, each divided into two generator subsystems, and performed in parallel on two computers. The overall execution time of centralized and distributed simulations under the same conditions was recorded. Figure 9Compared to centralized simulation, applying a distributed architecture to a four-machine, two-region system does not offer a time advantage. This is because the four-machine, two-region system is relatively small, and the time required to solve the system's differential-algebraic equations is very short. The additional time incurred by the distributed architecture, such as data transmission and waiting between subsystems, outweighs the time saved by parallel computing. As the system size increases, the number of differential-algebraic equations increases, and the time required for solving them increases. Correspondingly, after dividing the system into subsystems, the time saved by parallel computing gradually exceeds the additional time added by the distributed architecture. Compared to centralized simulation, in the improved New England 10-machine, 39-node system, using a distributed architecture on two machines can reduce simulation time by 10.65%; while in the larger WECC 29-machine, 179-node system, the distributed architecture further reduces simulation time by 48.73%, demonstrating a significant efficiency improvement. As the system size increases, the advantage of the distributed simulation architecture in reducing simulation time becomes more prominent. It is reasonable to infer that when applied to power systems with a larger node scale, the distributed simulation architecture is expected to achieve better time performance, providing stronger technical support for improving the simulation efficiency of large-scale power systems.

[0093] Therefore, the distributed simulation architecture has good scalability and can effectively improve the simulation efficiency of power systems while ensuring the accuracy of the results.

[0094] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A distributed simulation method for power systems based on data distribution services, characterized in that, Includes the following steps: The power system is divided into n+1 subsystems, including n generator subsystems and 1 network subsystem. A data distribution service is used to design communication interfaces to realize data exchange between subsystems. The nodes of the n+1 subsystems are mapped to at most n+1 devices. The subsystems are jointly simulated through the devices, and the devices can interact with each other. Parallel computation will be performed between the generator subsystems, while the generator subsystem and the network subsystem will be solved alternately; Subsystem partitioning is based on time-consumption balance. First, based on the differences in the system's physical topology and dynamic characteristics, the system is initially divided into subsystems. There are three basic subsystems, and the number of node devices that can participate in the simulation calculation is denoted as . When satisfied In this case, a device is allocated to each subsystem, and the subsystems are simulated in parallel. At that time, The subsystems are integrated into Group, make Total simulation time of subsystems on the group equipment Similar, if the groups are related If the gap remains significant, then integrate some subsystems and recalculate communication time and other parameters. and Through multiple iterations, the time consumption difference between groups is gradually reduced, ultimately minimizing the total simulation time on the K groups of devices. Approaching equilibrium.

2. The method according to claim 1, characterized in that, The joint simulation of the subsystem via the device includes: Get the start signal; Each subsystem receives data; According to the preset simulation time step, a step signal is sent to control the timing of subsystem simulation calculation and data exchange until all subsystems complete the preset simulation time. Each subsystem sends data; Obtain the stop signal.

3. The method according to claim 2, characterized in that, The step of sending step signals according to a preset simulation time step to control the timing of subsystem simulation calculations and data exchange includes: For two subsystems with different simulation time steps, the larger simulation time step is used. The corresponding subsystem runs one simulation step, and the simulation time step is a small value. The corresponding subsystem does not stop after completing one simulation step, but continues to run until the total time of its multi-step simulation equals [the required time]. At this point, both subsystems with the same simulation time step have completed the simulation for the current round.

4. The method according to claim 2, characterized in that, The network subsystem matches the maximum simulation time step of all generator subsystems.

5. The method according to claim 2, characterized in that, Before each subsystem begins simulation at each step, the currently received data is detected in real time to verify its integrity. A new round of simulation is triggered only when the data meets the simulation calculation requirements. If the data verification fails, the subsystem will automatically enter a blocking waiting state until updated data is received, at which point the blocking will be lifted immediately, ensuring that the subsystem always continues simulation calculation based on the latest valid data.

6. The method according to claim 1, characterized in that, At the data storage and synchronization level, the core data of each simulation subsystem model is stored on the local device. Only after each simulation step is completed, the interaction parameters are synchronized to the associated subsystems of other devices in real time through the coupled communication interface.

7. The method according to claim 1, characterized in that, The method for data exchange between subsystems includes the following steps: Within a subsystem that requires data interaction, publishers and subscribers are created, wherein the publisher is the data sender and the subscriber is the data receiver; Group the data to be transmitted according to parameter characteristics and register the data topics; After the publisher determines the topic of the data to be published, a topic table is generated in the data domain; after the subscriber determines the topic of the data to be subscribed to, a topic table is generated in the data domain. The publisher publishes data within the data domain; The subscriber listens to see if it has subscribed to data by querying the published topic table and the data field. If it has not subscribed to data, it continues to listen. If it has subscribed to data, it retrieves the data from the data field.

8. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method as described in any one of claims 1 to 7.

Images