Parallel multi-rate simulation method based on time constant network division decoupling

Abstract

The invention discloses a parallel multi-rate simulation method based on time constant sub-network decoupling, which comprises the following steps of: performing sub-network according to different time constants of electrical equipment of a power system, extracting sub-network simulation information such as a sub-network node admittance matrix and a historical current source column vector, and interactively decoupling each sub-network; according to the method, the state variables of the sub-networks can be independently solved, meanwhile, a parallel multi-rate simulation framework is constructed for the data interaction opportunity problem of the sub-networks after decoupling, and different time constant networks adopt different time rates for parallel simulation, that is, the sub-networks with different time constants propel the simulation process in parallel with different simulation step lengths. According to the method, the calculation complexity is reduced and the unnecessary calculation expenditure is reduced through the two aspects of sub-network dimension reduction and parallel multi-rate, the calculation resources are fully utilized, and the simulation efficiency of the electromagnetic transient state is improved on the premise of ensuring the simulation precision.

Description

Technical Field

[0001] This invention relates to the field of power simulation technology, specifically a parallel multi-rate simulation method based on time constant network decoupling. Background Technology

[0002] Under the national strategy of new power systems, large-scale wind and solar power are being integrated into the grid. Due to the fluctuating nature of these renewable energy sources, the electricity generated is typically processed by power electronic devices before being fed into the grid. However, power electronic converters lack the rotor inertia of traditional generator sets, posing a significant challenge to system frequency stability. Furthermore, due to differing control strategies, the massive number of connected power electronic devices are coupled and interact with each other, easily leading to complex electromagnetic transient processes such as harmonic oscillations. This significantly impacts the safety of electrical equipment and the stability of the system. Electromagnetic transient simulation of the power system is a crucial means of understanding these complex transient processes.

[0003] However, electromagnetic transient simulation of a large number of power electronic devices is not easy. The switching transistors in power electronic devices change at high frequencies, which means that if we want to focus on transient high-frequency components, we need to simulate them with extremely small step sizes. The constantly changing switching states mean that we need to invert the high-dimensional matrix at each time step, which puts a lot of pressure on modern computing devices. For real-time simulation, a lot of computing resources are needed, and for offline simulation, the simulation time is unacceptable.

[0004] To improve simulation efficiency, the main method used in engineering is to decouple the system and solve the subnets independently, thereby reducing computational complexity. Simultaneously, the relatively independent subnets after decoupling can be simulated in parallel, eliminating serial delays in the solution process and significantly improving efficiency. However, the time constants of different regions in power electronic networks vary considerably. Using only single-rate simulation methods cannot reflect the differences in dynamic characteristics of each subnet. Simulating subnets with large time constants using small step sizes increases unnecessary computational overhead. Therefore, engineering practices design multi-rate simulation processes based on the decoupling method, simulating subnets at different time rates and selecting the simulation step size according to actual needs, thus fully utilizing computational resources. Specific examples include the application of parallel multi-rate algorithms to methods such as transmission line decomposition, delay insertion, and multi-region Thevenin equivalence.

[0005] However, the aforementioned parallel multi-rate methods suffer from problems such as numerical oscillations caused by switching, poor stability, and loss of simulation accuracy. Because regions with significant time constant differences still exist within the decoupled network, the subnets tend to use small-step simulations to accommodate regions with smaller time constants, making it difficult to fully leverage the advantages of parallel multi-rate methods. These problems arise because the aforementioned parallel multi-rate methods are modifications of existing network decoupling schemes, and the original network decoupling methods are not fully adaptable to parallel multi-rate algorithms. Summary of the Invention

[0006] The purpose of this invention is to provide a parallel multi-rate simulation method based on time constant network decoupling, comprising the following steps:

[0007] Step 1: Divide the power system into different electrical devices and form multiple subnets;

[0008] Step 2: Each subnet extracts network information and performs interactive decoupling of network information;

[0009] The extraction and interaction of simulation information are decoupled from each other;

[0010] Step 3: Design the timing of subnet data interaction to enable parallel solution of each subnet, and construct a parallel multi-rate simulation framework;

[0011] Step 4: Perform step size transformation of subnet state variables at different time rates;

[0012] Step 5: Use the parallel multi-rate simulation framework to complete the parallel multi-rate electromagnetic transient simulation of the power system.

[0013] Furthermore, in step 1, the different electrical devices in the power system are divided into separate networks according to the magnitude of the time constant.

[0014] Furthermore, the resulting subnet is divided into a fast-changing subnet and a slow-changing subnet; in the fast-changing subnet, the power electronic switching frequency is greater than 1kHz, and the duration of the transient process and the time constant are less than 1ms; in the slow-changing subnet, the switching frequency is less than 1kHz, and the duration of the transient process and the time constant are greater than 1ms.

[0015] Subnets with transient process durations greater than a preset time threshold are classified as fast-changing subnets, while subnets with transient process durations less than or equal to the preset time threshold are classified as slow-changing subnets.

[0016] Furthermore, the network information includes the subnet node admittance matrix and the historical current source column vector.

[0017] Furthermore, in step 2, each subnet extracts network information to the connection nodes between subnets through matrix compression technology, and then broadcasts and matches the subnet information to achieve interactive decoupling of network information.

[0018] Furthermore, the steps for designing the timing of subnet data interaction are as follows: assign different simulation step sizes to fast-changing and slow-changing subnets so that the solutions of each subnet can be performed in parallel.

[0019] Furthermore, the simulation step size of the slowly varying subnet is greater than that of the rapidly varying subnet.

[0020] Furthermore, in step 4, performing step size conversion for state variables of subnets with different time rates means: performing step size conversion on the nodal admittance matrix elements extracted from the slow-changing subnet and the historical current source column vector elements to unify the step size of the nodal admittance matrix elements and the historical current source column vector elements of the slow-changing subnet and the fast-changing subnet.

[0021] Transformation of nodal admittance matrix elements refers to the explicit transformation of element conductance based on the step size;

[0022] Step-size transformation of the column vector elements of historical current sources refers to substituting the changed conductance into the historical current source solution formula to obtain the historical current source value after step-size transformation.

[0023] Furthermore, in step 5), when completing the parallel multi-rate electromagnetic transient simulation of the power system, after each step is solved, the simulation time is accumulated by the step size, and the fast and slow subnet states are solved synchronously at a fixed time rate.

[0024] Furthermore, when the simulation time for both the fast-changing subnet and the slow-changing subnet reaches the preset simulation time, the simulation ends, and the parallel multi-rate simulation of the power system is completed.

[0025] The technical effects of this invention are undeniable. This invention is more adaptable to parallel multi-rate algorithms and avoids the problem that after decoupling using traditional techniques, there are still regions with large differences in time constants within the subnet, which means that the subnet still needs to use small step size simulation to accommodate regions with small time constants. This invention can give full play to the advantages of parallel multi-rate and improve the efficiency of electromagnetic transient simulation and the utilization of computing resources. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of a power electronic converter testing system;

[0027] Figure 2 This is a schematic diagram of the interaction for extracting network information.

[0028] Figure 3 This is a schematic diagram of a serial multi-rate process;

[0029] Figure 4 This is a schematic diagram of a parallel multi-rate process;

[0030] Figure 5 To test the low-voltage DC side voltage waveform of the system;

[0031] Figure 6 To test the voltage waveform of the high voltage DC side of the system. Detailed Implementation

[0032] The present invention will be further described below with reference to embodiments, but it should not be construed that the scope of the present invention is limited to the following embodiments. Various substitutions and modifications made based on ordinary technical knowledge and common practices in the art without departing from the above-described technical concept of the present invention should be included within the scope of protection of the present invention.

[0033] Example 1:

[0034] See Figures 1 to 6 A parallel multi-rate simulation method based on time constant network decoupling includes the following steps:

[0035] Step 1: Divide the power system into different electrical devices and form multiple subnets;

[0036] Step 2: Each subnet extracts network information and performs interactive decoupling of network information;

[0037] The extraction and interaction of simulation information are decoupled from each other;

[0038] Step 3: Design the timing of subnet data interaction to enable parallel solution of each subnet, and construct a parallel multi-rate simulation framework;

[0039] Step 4: Perform step size transformation of subnet state variables at different time rates;

[0040] Step 5: Use the parallel multi-rate simulation framework to complete the parallel multi-rate electromagnetic transient simulation of the power system.

[0041] Example 2:

[0042] A parallel multi-rate simulation method based on time constant network decoupling is proposed. The technical content is the same as in Example 1. Further, in step 1, the different electrical devices in the power system are divided into networks according to the size of the time constant.

[0043] Example 3:

[0044] A parallel multi-rate simulation method based on time constant network decoupling, with the same technical content as any one of Embodiments 1-2, further wherein the formed subnet is divided into a fast-changing subnet and a slow-changing subnet; in the fast-changing subnet, the power electronic switching frequency is greater than 1kHz, and the transient process duration and time constant are less than 1ms; in the slow-changing subnet, the switching frequency is less than 1kHz, and the transient process duration and time constant are greater than 1ms.

[0045] Subnets with transient process durations greater than a preset time threshold are classified as fast-changing subnets, while subnets with transient process durations less than or equal to the preset time threshold are classified as slow-changing subnets.

[0046] Example 4:

[0047] A parallel multi-rate simulation method based on time constant network decoupling, with the same technical content as any one of embodiments 1-3, further wherein the network information includes the subnet node admittance matrix and the historical current source column vector.

[0048] Example 5:

[0049] A parallel multi-rate simulation method based on time constant network decoupling, with the same technical content as any one of embodiments 1-4, further wherein in step 2, each subnet extracts network information to the connection nodes between subnets through matrix compression technology, and then broadcasts and matches the subnet information to achieve interactive decoupling of network information.

[0050] Example 6:

[0051] A parallel multi-rate simulation method based on time constant network decoupling, with the same technical content as any one of embodiments 1-5, further wherein the step of designing the timing of subnet data interaction is as follows: assigning different simulation step sizes to fast-changing subnets and slow-changing subnets so that the solution of each subnet can be performed in parallel.

[0052] Example 7:

[0053] A parallel multi-rate simulation method based on time constant network decoupling, with the same technical content as any one of embodiments 1-6, further wherein the simulation step size of the slow-changing subnet is greater than that of the fast-changing subnet.

[0054] Example 8:

[0055] A parallel multi-rate simulation method based on time constant network decoupling, with the same technical content as any one of embodiments 1-7, further, in step 4, performing step size conversion of state variables of subnets with different time rates means: performing step size conversion on the node admittance matrix elements extracted from the slow-changing subnet and the historical current source column vector elements, so that the step size of the node admittance matrix elements and the historical current source column vector elements of the slow-changing subnet and the fast-changing subnet is unified;

[0056] Transformation of nodal admittance matrix elements refers to the explicit transformation of element conductance based on the step size;

[0057] Step-size transformation of the column vector elements of historical current sources refers to substituting the changed conductance into the historical current source solution formula to obtain the historical current source value after step-size transformation.

[0058] Example 9:

[0059] A parallel multi-rate simulation method based on time constant network decoupling, with the same technical content as any one of embodiments 1-8, further wherein, in step 5), when completing the parallel multi-rate electromagnetic transient simulation of the power system, after the solution of each step size is completed, the simulation time is accumulated by the step size, and the fast and slow sub-network states are solved synchronously at a fixed time rate.

[0060] Example 10:

[0061] A parallel multi-rate simulation method based on time constant network decoupling, with the same technical content as any one of embodiments 1-9, further wherein the simulation ends when the simulation time of both the fast-changing subnet and the slow-changing subnet reaches the preset simulation time, and the parallel multi-rate simulation of the power system is completed.

[0062] Example 11:

[0063] A parallel multi-rate simulation method based on time constant-based network decoupling fully utilizes computational resources and improves simulation efficiency. Unlike existing simulation techniques that first establish a decoupling method and then construct a multi-rate scheme, this invention reverse-engineers the network decoupling method with parallel multi-rate as the target. This allows each sub-network after decoupling to naturally adapt to parallel simulation at different time rates, avoiding the problem that regions with large time constant differences still exist within sub-networks after decoupling using traditional techniques, thus preventing the full realization of the advantages of parallel multi-rate simulation. Specifically, it includes:

[0064] Step 1: Divide the electrical equipment in different parts of the power system into separate networks according to the magnitude of the time constant;

[0065] Step 2: Each subnet decouples itself from the others through the extraction and interaction of simulation information;

[0066] Step 3: Construct a parallel multi-rate simulation framework for subnet data interaction timing;

[0067] Step 4: Step size transformation of subnet state variables at different time rates;

[0068] Step 5: Complete the parallel multi-rate simulation of the power system.

[0069] Step 1, which involves dividing the network based on the time constant, specifically involves:

[0070] Power systems consist of networks with different time constants. Using different simulation step sizes for networks with different time constants can greatly save computational resources. The system is classified according to its time constant, forming different subnets. Subnets containing power electronic switches and those with smaller time constants are classified as fast-changing subnets, while the remaining networks are classified as slow-changing subnets.

[0071] Step 2, subnet information extraction, interaction, and decoupling, specifically involves:

[0072] The node admittance matrix and historical current source column vectors of a subnet represent all the simulation information of that subnet. By using matrix compression techniques to extract network information to the connection nodes between subnets, information exchange between subnets becomes possible. The information from each subnet interacts through the superposition of matrices and vectors. After information exchange, each subnet gains the ability to observe the external system, thus achieving decoupling between subnets.

[0073] The parallel multi-rate simulation framework described in step 3 is as follows:

[0074] By designing the timing of data interaction between subnets, the solutions to each subnet can be solved in parallel, improving simulation efficiency. Since network decoupling is already performed based on the network time constant, subnets can be easily solved with different simulation step sizes. Fast-changing subnets are simulated with small step sizes for finer detail, reflecting high-frequency characteristics, while slow-changing subnets are simulated with large step sizes, reducing unnecessary computational overhead and alleviating computational pressure. Information between subnets of different rates needs to be selected according to certain principles; for example, data exchanged from fast-changing subnets to slow-changing subnets needs to be averaged before transmission.

[0075] The specific steps of the state quantity step size transition in step 4 are as follows:

[0076] After the electrical component is discretized, it consists of conductance and historical current source. These two state variables are closely related to the simulation step size. The component conductance can be explicitly converted according to the step size, while the conversion of the historical current source is achieved by utilizing the characteristic that the external characteristics of the electrical component's terminal voltage and inflow current remain unchanged after the time step is solved.

[0077] Step 5 involves accumulating the simulation time by step size after each step is completed, and simultaneously solving the fast and slow subnet states at a fixed time rate. When both fast and slow subnets simultaneously meet the simulation termination condition, the parallel multi-rate simulation of the power system is completed.

[0078] Example 12:

[0079] A parallel multi-rate simulation method based on time constant network decoupling is presented below. The specific implementation steps of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0080] Step 1: Divide the electrical equipment in different parts of the power system into separate networks according to the size of the time constant.

[0081] Networks with constantly changing states, networks with small time constants (i.e., long transient process durations), and networks whose high-frequency characteristics need to be observed are considered fast-changing subnets, while other networks are considered slow-changing subnets.

[0082] Figure 1 This is a schematic diagram of a power electronic converter test system, specifically a dual active bridge DC-DC transformer system. In this system, the H-bridge converts DC to AC via rapid switching by power electronic switches. To reduce transformer size, this AC frequency is relatively high. To observe the high-frequency components in the high-frequency transformer, a small-step simulation is required; therefore, the high-frequency transformer and the H-bridges on both sides are considered as a fast-changing subnetwork. The electrical quantities of the DC source and DC load do not change drastically, so they are considered as a slow-changing subnetwork.

[0083] Step 2: Each subnet decouples itself from the others through the extraction and interaction of simulation information.

[0084] The node admittance matrix and historical current source column vectors of a subnet represent all the simulation information of that subnet. By using matrix compression techniques, network information can be extracted to the connection nodes between subnets, enabling the broadcasting and matching of subnet information. This is because network information exchange requires common components, and the connection nodes between subnets are the only shared part; therefore, the extracted network information will have the same dimension as the connection nodes. Network information can be exchanged through matrix overlay. After this interaction, each subnet gains the ability to observe the external system, thus decoupling the subnets.

[0085] Figure 2 This paper illustrates the complete process of extracting and interacting network information, showing how information from subnet A is extracted and then shared with subnet B. Subnets A and B are connected via α-dimensional nodes. Using Ward's isometry, all information from subnet A is extracted to the α-dimensional connection nodes. The extracted information from subnet A is then superimposed onto the corresponding α-dimensional boundary nodes of subnet B, enabling subnet B to fully observe subnet A. Solving subnet B no longer depends on subnet A. The observation of subnet B from subnet A follows the same logic, thus achieving decoupling between subnets.

[0086] Step 3: Construct a parallel multi-rate simulation framework for subnet data interaction timing;

[0087] Based on subnet decoupling, the solution of each subnet is made in parallel by designing the timing of subnet data interaction. Different types of subnets are solved with different simulation step sizes. When solving the rapidly changing subnet, how to account for the contribution of the slowly changing subnet, and how to reflect the contribution of the rapidly changing subnet in the large step size after the solution of the rapidly changing subnet is a problem that the multi-rate framework needs to solve.

[0088] The information exchanged between subnets includes the node admittance matrix and the historical current source column vector. For the node admittance matrix, since there are no power electronic switches in the slow-changing subnet, the node admittance matrix will not change. However, the node admittance matrix of the fast-changing subnet contains switches, and the node admittance matrix will change continuously when the switches are modeled using the binary resistor method. To solve this problem, this invention applies the substitution theorem and uses current sources to equivalently replace the influence of switches on the subnet.

[0089] The historical current source column vectors passed from the slow-changing subnet to the fast-changing subnet are not processed; that is, within a large step size, it is assumed that each small step size of the fast-changing subnet uses the same historical current source information as the slow-changing subnet. For historical current sources passed from the fast-changing subnet to the slow-changing subnet, since the fast-changing subnet generates more historical current source information due to its higher solution frequency, the method used is to average the historical current sources of each small step size of the fast-changing subnet and then pass them back to the fast-changing subnet.

[0090] The following section will first introduce the multi-rate serial process, and then gradually transition to the multi-rate parallel process.

[0091] The fast-changing subnet has a step size of Δt, and the slow-changing subnet has a step size of ΔT, where nΔt = ΔT. Figure 3 This is a serial multi-rate process.

[0092] Step 1: Starting at time t, after the information exchange between the fast-changing subnet and the slow-changing subnet is completed, the slow-changing subnet can solve the system with a time step of ΔT to obtain the voltage of all nodes in the slow-changing subnet at time t+ΔT.

[0093] Step 2: Extract network information from the slow-changing subnet to the fast-changing subnet.

[0094] Step 3: After receiving the information from the slow-changing subnet, the fast-changing subnet interpolates the information of the slow-changing subnet at each small step size based on the historical information of the slow-changing subnet. Simulation starts from time t with a time step of Δt, and after n steps, it reaches t+ΔT, which is aligned with the simulation time of the slow-changing subnet.

[0095] Step 4: The topology of the fast-changing subnet may change multiple times within a time step ΔT. An averaging method is used to average the network information at each time step Δt to reflect the comprehensive impact of the fast-changing subnet on the slow-changing subnet. Topology changes indicate changes in switch states. In this invention, current sources are used to replace switches to simulate topology changes. This ensures that the conductance information remains constant at each time step Δt, and the topology change is only reflected in the current sources; therefore, only the historical current source information needs to be averaged.

[0096] As can be seen in this process, the slow-variable subnet and the fast-variable subnet are solved sequentially. Within a complete step, after the fast-variable subnet advances by a large step, the slow-variable subnet starts from the original time and advances by n small steps to achieve synchronization. The large and small step systems exchange data twice through interpolation and averaging processes, which makes the simulation time relatively long.

[0097] Figure 4 This is a schematic diagram of a parallel multi-rate process. For parallel multi-rate processes, the following applies:

[0098] Taking the simulation at time t as an example, network information is extracted for the slowly varying subnet, while the rapidly varying subnet follows the same procedure as in step four of the serial process, forming averaged network information. After obtaining the information of the other subnet, the two subnets can be solved in parallel. Unlike the serial process, due to the parallel solution, the long-history current source information of the slowly varying subnet is not obtained when solving the rapidly varying subnet. If interpolation is performed at this time, it will become extrapolation prediction, which poses a stability problem for the simulation. Therefore, the information passed from the slowly varying subnet to the rapidly varying subnet is no longer interpolated. That is, under a large step size, the same slowly varying subnet information is used for each small step size of the rapidly varying subnet and does not change.

[0099] Compared to the serial process, the parallel process solves both the slow-changing subnet and the fast-changing subnet simultaneously, greatly improving computational efficiency.

[0100] Step 4: Step size transformation of subnet state variables at different time rates.

[0101] After the electrical components are discretized, they consist of conductance and historical current sources. These two state variables are closely related to the simulation step size. The component conductance can be explicitly converted according to the step size. Since the external characteristics of the branch remain unchanged, the changed conductance can be substituted into the historical current source solution formula to obtain the historical current source value after the step size conversion.

[0102] Step 5: Complete the parallel multi-rate simulation of the power system.

[0103] It is worth noting that, due to the different step sizes between subnets, network information cannot be directly exchanged. A step size conversion is required between the node admittance matrix elements extracted from the slow-changing subnet and the historical current source column vector elements. After each step is solved, the state of the power system at that time step is known. The simulation time is accumulated by the step size. Since the electromagnetic transient simulation step size is small, the system state, which changes approximately continuously over time, can be obtained. The states of the fast-changing and slow-changing subnets advance synchronously at a certain time ratio. When both the fast and slow subnets simultaneously meet the simulation termination condition, the parallel multi-rate simulation of the power system is completed.

[0104] right Figure 1 The system shown is solved using a parallel multi-rate simulation method based on time constant network decoupling. The simulation results are compared and analyzed with PSCAD / EMTDC. A short circuit is set on the low-voltage DC side via a 0.005Ω resistor at 0.4s in the simulation. The short-circuit fault lasts for 2ms and is cleared, after which the system enters the fault recovery process until it reaches a steady state. The fast-changing subnet and the slow-changing subnet are simulated at a rate of 100:1. Figure 5 , Figure 6 The voltage waveforms on the high-voltage DC side and low-voltage DC side of the test system are shown respectively. Comparison shows that the relative error between the proposed method and the PSCAD method during transient periods is within 5%, accurately reflecting system characteristics. When the fast-changing subnet is simulated with a step size of 2.5 μs, the slow-changing subnet with a step size of 250 μs, and PSCAD with a step size of 2.5 μs, the parallel multi-rate simulation method based on time constant network decoupling has a speedup effect of more than 5 times compared to PSCAD / EMTDC, significantly improving simulation efficiency.

Claims

1. A parallel multi-rate simulation method based on time constant network decoupling, characterized in that, Includes the following steps: Step 1: Divide the power system into different electrical devices and form multiple subnets; Step 2: Each subnet extracts network information and performs interactive decoupling of network information; The extraction and interaction of simulation information are decoupled from each other; Step 3: Design the timing of subnet data interaction to enable parallel solution of each subnet, and construct a parallel multi-rate simulation framework; Step 4: Perform step size transformation of subnet state variables at different time rates; Step 5: Use the parallel multi-rate simulation framework to complete the parallel multi-rate electromagnetic transient simulation of the power system.

2. The parallel multi-rate simulation method based on time constant network decoupling according to claim 1, characterized in that, In step 1, the different electrical devices in the power system are divided into sub-networks according to the size of the time constant.

3. The parallel multi-rate simulation method based on time constant network decoupling according to claim 2, characterized in that, The resulting subnet is divided into a fast-changing subnet and a slow-changing subnet. In the fast-changing subnet, the power electronic switching frequency is greater than 1 kHz, and the duration of the transient process and the time constant are less than 1 ms. In the slow-changing subnet, the switching frequency is less than 1 kHz, and the duration of the transient process and the time constant are greater than 1 ms. Subnets with transient process durations greater than a preset time threshold are classified as fast-changing subnets, while subnets with transient process durations less than or equal to the preset time threshold are classified as slow-changing subnets.

4. The parallel multi-rate simulation method based on time constant network decoupling according to claim 1, characterized in that, The network information includes the admittance matrix of subnet nodes and the column vector of historical current sources.

5. The parallel multi-rate simulation method based on time constant network decoupling according to claim 1, characterized in that, In step 2, each subnet extracts network information to the connection nodes between subnets through matrix compression technology, and then broadcasts and matches the subnet information to achieve interactive decoupling of network information.

6. The parallel multi-rate simulation method based on time constant network decoupling according to claim 1, characterized in that, The steps for designing the timing of subnet data interaction are as follows: assign different simulation step sizes to fast-changing and slow-changing subnets so that the solutions of each subnet can be performed in parallel.

7. The parallel multi-rate simulation method based on time constant network decoupling according to claim 6, characterized in that, The simulation step size of the slow-changing subnet is greater than that of the fast-changing subnet.

8. The parallel multi-rate simulation method based on time constant network decoupling according to claim 1, characterized in that, In step 4, the step size conversion of the state variables of the subnets with different time rates refers to the step size conversion of the nodal admittance matrix elements extracted from the slow-changing subnet and the historical current source column vector elements, so that the step size of the nodal admittance matrix elements and the historical current source column vector elements of the slow-changing subnet and the fast-changing subnet is unified. Transformation of nodal admittance matrix elements refers to the explicit transformation of element conductance based on the step size; Step-size transformation of the column vector elements of historical current sources refers to substituting the changed conductance into the historical current source solution formula to obtain the historical current source value after step-size transformation.

9. The parallel multi-rate simulation method based on time constant network decoupling according to claim 1, characterized in that, In step 5), when completing the parallel multi-rate electromagnetic transient simulation of the power system, after each step is solved, the simulation time is accumulated by the step size, and the fast and slow subnet states are solved synchronously at a fixed time rate.

10. A parallel multi-rate simulation method based on time constant network decoupling according to claim 9, characterized in that, The simulation ends when the simulation time for both the fast-changing subnet and the slow-changing subnet reaches the preset simulation time, and the parallel multi-rate simulation of the power system is completed.

Images