A device-level accurate equivalent modeling and high-efficiency simulation method for dual active bridge converter

Abstract

The application discloses a kind of double active bridge converter device-level accurate equivalent modeling and high-efficiency simulation method, comprising the following steps: step 1) constructing double active bridge converter circuit topology;Step 2) constructing double active bridge converter IGBT high-order nonlinear model;Step 3) IGBT high-order nonlinear model is equivalent to single-port Norton equivalent circuit;Step 4) single-port Norton equivalent circuit is embedded in double active bridge converter circuit topology, and second cut-set contraction equivalent simplification is carried out, to obtain the double-port equivalent circuit model of the DAB converter;Step 5) establish implicit iterative electromagnetic transient simulation framework, electromagnetic transient simulation framework is solved iteratively in parallel computing mode, to obtain the nanosecond switching dynamic response of IGBT, so as to realize the device-level high-speed simulation of DAB converter.The simulation waveform of the model of the application is in good agreement with the measured results, can fully reflect the internal transient characteristics of DAB converter, and is greatly superior to traditional simulation method in efficiency.

Description

Technical Field

[0001] This invention relates to the field of power simulation technology, specifically a method for accurate equivalent modeling and efficient simulation of dual active bridge converters at the device level. Background Technology

[0002] Dual active bridge (DAB) DC-DC converters have been widely used in high-voltage DC transmission and medium- and low-voltage DC distribution due to their advantages such as high power density, support for bidirectional power transfer, and electrical isolation. They are particularly prevalent in modular power electronic equipment such as cascaded solid-state transformers. DAB converters typically contain multiple high-speed switching power electronic devices (such as multiple IGBT power transistors and anti-parallel diodes). The switching transients during operation can induce complex phenomena such as high-order harmonics and high-frequency oscillations, leading to increased voltage and current stress on the devices, increased switching losses, and in severe cases, damage to the devices due to overvoltage or overcurrent. Therefore, even in system-level studies, it is necessary to incorporate the analysis of the switching dynamic characteristics of power electronic devices.

[0003] Electromagnetic transient (EMT) modeling and simulation are important tools for studying the dynamic behavior of power electronic systems, including those involving digital switching (DAB). Detailed simulation at the device level can accurately assess the losses and stresses of switching devices, providing a basis for topology optimization, device selection, and control strategy formulation. However, due to the complexity of device-level models, direct system-level device-level simulation is extremely inefficient. Currently, there are three main types of switching device models used for DAB transient simulation: ideal models equate switches to binary elements driven by control signals (such as binary resistors and inductors), without depicting the transient process of the turn-on signal. These models offer fast simulation speeds but can only be used for macroscopic simulations at the device or system level; behavioral models fit and model the macroscopic voltage and current characteristics of the switching transistor ports, reflecting certain dynamic characteristics of the switches at the port level. Some improved behavioral models support efficient simulations with small step sizes in the microsecond or even nanosecond range, balancing computational speed with a certain level of accuracy. However, because they do not describe the internal physical processes of the device, errors still exist in the fast transient behavior within nanoseconds of the number of switching transistors. Physical models are established based on physical mechanisms such as the movement of charge carriers within the device, and typically use equivalent circuits and nonlinear current sources to represent the device's physical differential equations (e.g., the Hefner model and the Karus model). These models can most accurately describe the physical dynamics of switching devices and have the highest accuracy. However, the model structure is complex, requiring multiple iterations of the nonlinear equations to be solved within each simulation step, resulting in extremely low simulation efficiency. Taking the Hefner physical model as an example, an IGBT / diode switching pair includes approximately 8 electrical nodes. When this model is applied to a DAB converter containing multiple switching transistors, the total number of electrical nodes jumps from the original 9 to approximately 57. Considering that the computational complexity of solving electromagnetic transients increases cubically with the number of nodes, the significant increase in the number of nodes will lead to a reduction in simulation speed by several orders of magnitude (approximately 254 times slower). Furthermore, accurate device-level simulation requires using step sizes much smaller than those for system-level simulation to characterize the nanosecond-level dynamics of the IGBT, and to iterate and solve the device's nonlinear differential equations multiple times within each step. If the step size is reduced from microseconds (e.g., 2µs) to nanoseconds (e.g., 10ns), and 10 calculations are performed per step, the overall simulation speed will decrease by approximately 500,000 times. Therefore, directly using detailed physical models for device-level simulation of DAB converters is computationally unacceptably time-consuming.

[0004] To improve the efficiency of device-level simulation, existing technologies typically employ a combination of network decoupling and parallel computing. On one hand, complex power electronic networks are segmented and decoupled according to regions or characteristics, allowing each sub-network to be solved independently, thus reducing the scale of each solution. On the other hand, multi-core CPUs or GPUs are used to accelerate simulation computation. While these methods improve simulation speed to some extent, they also have significant drawbacks: network decoupling may introduce additional numerical errors and oscillation risks, requiring a sacrifice of some simulation accuracy; while the acceleration effect of parallel computing is limited by hardware performance, with significant marginal effects when facing extremely complex problems, making it difficult to achieve unlimited linear acceleration. Furthermore, some studies have attempted to transfer network equivalent methods from large-scale power system simulation to device-level models, such as network tearing, Ward's equivalent, and multi-port Thevenin's equivalent. However, due to the smaller step size, stronger rigidity, and implicit iteration characteristics of device-level simulation, the direct application of these general equivalent methods is very difficult and requires targeted improvements. In summary, existing technologies struggle to simultaneously meet the high accuracy and high efficiency requirements of DAB converter device-level simulation. Accelerating simulation computation while preserving device details is a pressing technical problem to be solved in this field. Summary of the Invention

[0005] The purpose of this invention is to provide a method for accurate equivalent modeling and efficient simulation of dual active bridge converters at the device level, comprising the following steps:

[0006] Step 1) Construct the dual active bridge converter circuit topology;

[0007] Step 2) Construct a high-order nonlinear model of the IGBT in the dual active bridge converter;

[0008] Step 3) Using the cut set contraction dimensionality reduction equivalent method, the IGBT high-order nonlinear model is equivalent to a single-port Norton equivalent circuit;

[0009] Step 4) Embed the single-port Norton equivalent circuit into the dual active bridge converter circuit topology, and perform a second cut-set contraction simplification on the embedded dual active bridge converter circuit topology to obtain the dual-port equivalent circuit model characterizing the DAB converter.

[0010] Step 5) Based on the two-port equivalent circuit model, an implicit iterative electromagnetic transient simulation framework is established. The electromagnetic transient simulation framework is iteratively solved in parallel computing mode to obtain the nanosecond-level switching dynamic response of the IGBT, thereby realizing the device-level high-speed simulation of the DAB converter.

[0011] Furthermore, in step 1), the dual active bridge converter circuit structure includes the H-bridge primary circuit, the high-frequency transformer, and the H-bridge secondary circuit;

[0012] Both the primary and secondary circuits of the H-bridge include multiple IGBTs; each IGBT has an anti-parallel diode FWD.

[0013] Furthermore, in step 2), the steps for constructing the higher-order nonlinear model of the IGBT include:

[0014] Based on the physical characteristics of IGBT, the internal structure of IGBT is equivalent to a Darlington cascade of MOSFET transistors and BJT transistors, thereby constructing a high-order nonlinear model of IGBT containing multiple internal nodes.

[0015] Furthermore, in step 3), the single-port Norton equivalent circuit includes equivalent admittance and equivalent current source.

[0016] Furthermore, in step 3), the step of equipping the IGBT high-order nonlinear model with a single-port Norton equivalent circuit includes:

[0017] Step 3.1) Discretize the high-order nonlinear model of the IGBT to obtain the original node admittance matrix;

[0018] Step 3.2) Decompose the original node admittance matrix into symmetric components and oblique symmetric components, and correct the equivalent injected current source according to the oblique symmetric components to obtain a symmetric equivalent admittance matrix and a corrected current source, thereby transforming the IGBT high-order nonlinear model into a single-port Norton equivalent circuit with a symmetric node admittance matrix and an accompanying current source.

[0019] Furthermore, in step 4), the second cut-set contraction and simplification of the embedded dual active bridge converter circuit topology includes:

[0020] Step 4.1) Construct a directed graph that reflects the internal network structure of the DAB converter;

[0021] Step 4.2) Select one internal node as the reference node and at least one internal node as the external interface node to establish a cut set matrix;

[0022] Step 4.3) By partitioning nodes and performing circuit equivalence operations, the multiple nodes and branches contained within the DAB converter are simplified and reduced to obtain a two-port equivalent circuit model for connecting the DC input and output terminals of the DAB converter.

[0023] Furthermore, in step 4.3), when simplifying the multiple nodes and branches contained within the DAB converter, if a node is not directly connected to an external port, control signal, or measurement point, and its adjacent branches can be equivalently expressed by linear or nonlinear elements, then the node is included in the cut set for parallel elimination.

[0024] Circuit equivalent operations refer to transforming the directed graph of a DAB converter into a standard Norton structure through node rearrangement, matrix splitting, and cut-set transformation, while preserving the admittance G_eq and injected current I_eq of the external ports; G_eq=C t *G*C, I_eq=C t *J; where C is the cut set projection matrix, used to map internal nodes to port nodes; G is the node admittance matrix; and J is the associated current source matrix.

[0025] Furthermore, in step 5), the steps of simulating and solving the electromagnetic transient simulation framework using parallel computing include:

[0026] The steps for simulating and solving the electromagnetic transient simulation framework using parallel computing include:

[0027] In each simulation step, the node voltages and branch currents of the two-port equivalent circuit model are solved, and the historical state variables of the transformer, capacitor and IGBT models are updated after the step ends, until the preset convergence conditions are met.

[0028] Furthermore, the convergence condition is that the value of the state variable becomes less than a preset value or the maximum number of iterations is reached.

[0029] Specifically, if in two consecutive iterations, the voltage and current changes at all device ports are lower than a set threshold (e.g., 10⁻⁻⁴). 4 Up to 10⁻ 6 If the time interval is less than or equal to the order of magnitude of the given time, then the current step size is considered to have converged; otherwise, the iteration continues at the current time point until the condition is met or the maximum number of iterations is reached.

[0030] Furthermore, the electromagnetic transient simulation framework includes a transformer submodule, a capacitor submodule, and an IGBT device submodule;

[0031] When performing simulation on the electromagnetic transient simulation framework, the state updates and equivalent calculations of the transformer submodule, capacitor submodule, and IGBT device submodule are performed in parallel.

[0032] The technical effects of this invention are undeniable, and its beneficial effects are as follows:

[0033] 1. Through multi-level equivalent dimensionality reduction, the electrical network size of the DAB converter device-level model is significantly compressed, and the computational complexity is significantly reduced;

[0034] 2. With the implicit iterative solution and parallel computing framework, there is no global serial bottleneck caused by rigid small step size during the simulation process, which can make full use of the performance of multi-core hardware to achieve high-speed simulation.

[0035] This invention maintains nanosecond-level simulation accuracy for the dynamic characteristics of switching devices while increasing simulation speed to near the level of traditional ideal switch models, avoiding the additional errors and stability issues associated with traditional network decoupling acceleration methods. Actual test results show that the waveform accuracy of the model simulated by this invention is significantly better than simplified models such as ideal switches, while the simulation time is close to the latter. This achieves a comprehensive improvement in the accuracy and efficiency of electromagnetic transient simulation, and has significant engineering practical value for system-level simulation containing a large number of power electronic devices. Attached Figure Description

[0036] Figure 1 This is a diagram of the DAB converter topology.

[0037] Figure 2 This is a diagram of the equivalent circuit structure of an IGBT device.

[0038] Figure 3 Flowchart for IGBT model parameter extraction;

[0039] Figure 4 IGBT directed graph model and cut set structure;

[0040] Figure 5 This is the equivalent circuit diagram of a DAB circuit.

[0041] Figure 6 This is a DAB quadratic cut set contracted directed graph;

[0042] Figure 7 This is a flowchart of the parallel simulation process;

[0043] Figure 8 This is a comparison diagram of the primary and secondary waveforms of the transformer;

[0044] Figure 9 A comparison chart of IGBT voltage transients;

[0045] Figure 10 This is a comparison chart of the output voltage step response. Detailed Implementation

[0046] 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.

[0047] Example 1:

[0048] See Figures 1 to 10 A method for accurate equivalent modeling and efficient simulation of dual active bridge converters at the device level includes the following steps:

[0049] Step 1) Construct the dual active bridge converter circuit topology;

[0050] Step 2) Construct a high-order nonlinear model of the IGBT in the dual active bridge converter;

[0051] Step 3) Using the cut set contraction dimensionality reduction equivalent method, the IGBT high-order nonlinear model is equivalent to a single-port Norton equivalent circuit;

[0052] Step 4) Embed the single-port Norton equivalent circuit into the dual active bridge converter circuit topology, and perform a second cut-set contraction simplification on the embedded dual active bridge converter circuit topology to obtain the dual-port equivalent circuit model characterizing the DAB converter.

[0053] Step 5) Based on the two-port equivalent circuit model, an implicit iterative electromagnetic transient simulation framework is established. The electromagnetic transient simulation framework is iteratively solved in parallel computing mode to obtain the nanosecond-level switching dynamic response of the IGBT, thereby realizing the device-level high-speed simulation of the DAB converter.

[0054] Example 2:

[0055] See Figures 1 to 10 A method for accurate equivalent modeling and efficient simulation of a dual active bridge converter at the device level, with the same technical content as in Example 1, further, in step 1), the circuit structure of the dual active bridge converter includes the H-bridge primary circuit, the high-frequency transformer and the H-bridge secondary circuit.

[0056] Both the primary and secondary circuits of the H-bridge include multiple IGBTs; each IGBT has an anti-parallel diode FWD.

[0057] Example 3:

[0058] See Figures 1 to 10 A method for accurate equivalent modeling and efficient simulation of dual active bridge converters at the device level, with the same technical content as any one of Embodiments 1-2, further comprising the following steps in step 2) for constructing the high-order nonlinear model of the IGBT:

[0059] Based on the physical characteristics of IGBT, the internal structure of IGBT is equivalent to a Darlington cascade of MOSFET transistors and BJT transistors, thereby constructing a high-order nonlinear model of IGBT containing multiple internal nodes.

[0060] Example 4:

[0061] See Figures 1 to 10 A method for accurate equivalent modeling and efficient simulation of dual active bridge converters at the device level, with the same technical content as any one of embodiments 1-3. Further, in step 3), the single-port Norton equivalent circuit includes equivalent admittance and equivalent current source.

[0062] Example 5:

[0063] See Figures 1 to 10 A method for accurate equivalent modeling and efficient simulation of dual active bridge converters at the device level, with technical content identical to any one of embodiments 1-4, further comprising, in step 3), the step of equivalencing the IGBT high-order nonlinear model to a single-port Norton equivalent circuit includes:

[0064] Step 3.1) Discretize the high-order nonlinear model of the IGBT to obtain the original node admittance matrix;

[0065] Step 3.2) Decompose the original node admittance matrix into symmetric components and oblique symmetric components, and correct the equivalent injected current source according to the oblique symmetric components to obtain a symmetric equivalent admittance matrix and a corrected current source, thereby transforming the IGBT high-order nonlinear model into a single-port Norton equivalent circuit with a symmetric node admittance matrix and an accompanying current source.

[0066] Example 6:

[0067] See Figures 1 to 10 A method for accurate equivalent modeling and efficient simulation of a dual active bridge converter at the device level, with the same technical content as any one of embodiments 1-5, further comprising, in step 4), the step of performing a second cut-set contraction equivalent simplification on the embedded dual active bridge converter circuit topology, including:

[0068] Step 4.1) Construct a directed graph that reflects the internal network structure of the DAB converter;

[0069] Step 4.2) Select one internal node as the reference node and at least one internal node as the external interface node to establish a cut set matrix;

[0070] Step 4.3) By partitioning nodes and performing circuit equivalence operations, the multiple nodes and branches contained within the DAB converter are simplified and reduced to obtain a two-port equivalent circuit model for connecting the DC input and output terminals of the DAB converter.

[0071] Example 7:

[0072] See Figures 1 to 10 A method for accurate equivalent modeling and efficient simulation of dual active bridge converters at the device level, with the same technical content as any one of embodiments 1-6. Further, in step 4.3), when the multiple nodes and branches contained in the DAB converter are simplified by shrinking, if the node is not directly connected to an external port, control signal or measurement point, and its adjacent branch can be equivalently expressed by linear or nonlinear elements, then the node is included in the cut set for parallel elimination.

[0073] Circuit equivalent operations refer to transforming the directed graph of a DAB converter into a standard Norton structure through node rearrangement, matrix splitting, and cut-set transformation, while preserving the admittance G_eq and injected current I_eq of the external ports; G_eq=C t *G*C, I_eq=C t *J; where C is the cut set projection matrix, used to map internal nodes to port nodes; G is the node admittance matrix; and J is the associated current source matrix. C t It is the generalized inverse of C.

[0074] Example 8:

[0075] See Figures 1 to 10 A method for accurate equivalent modeling and efficient simulation of a dual active bridge converter at the device level, with technical content identical to any one of embodiments 1-7, further comprising, in step 5), the step of simulating and solving the electromagnetic transient simulation framework in a parallel computing manner, including:

[0076] The steps for simulating and solving the electromagnetic transient simulation framework using parallel computing include:

[0077] In each simulation step, the node voltages and branch currents of the two-port equivalent circuit model are solved, and the historical state variables of the transformer, capacitor and IGBT models are updated after the step ends, until the preset convergence conditions are met.

[0078] Example 9:

[0079] See Figures 1 to 10 A method for accurate equivalent modeling and efficient simulation of dual active bridge converters at the device level, with the same technical content as any one of embodiments 1-8, further wherein the convergence condition is that the value of the state variable becomes less than a preset value or reaches the maximum number of iterations.

[0080] Specifically, if in two consecutive iterations, the voltage and current changes at all device ports are lower than a set threshold (e.g., 10⁻⁻⁴). 4 Up to 10⁻ 6 If the time interval is less than or equal to the order of magnitude of the given time, then the current step size is considered to have converged; otherwise, the iteration continues at the current time point until the condition is met or the maximum number of iterations is reached.

[0081] Example 10:

[0082] See Figures 1 to 10 A method for accurate equivalent modeling and efficient simulation of a dual active bridge converter at the device level, with the same technical content as any one of embodiments 1-9. Further, the electromagnetic transient simulation framework includes a transformer submodule, a capacitor submodule, and an IGBT device submodule.

[0083] When performing simulation on the electromagnetic transient simulation framework, the state updates and equivalent calculations of the transformer submodule, capacitor submodule, and IGBT device submodule are performed in parallel.

[0084] Example 11:

[0085] See Figures 1 to 10 A method for accurate equivalent modeling and efficient simulation of dual active bridge converters at the device level, with technical content identical to any one of Examples 1-10, further comprising the electromagnetic transient simulation framework as follows: Figure 7 As shown, this framework is based on multi-layer equivalence and module partitioning, which transforms complex device-level networks (such as DAB converters) into smaller, more easily decoupled sub-modules through admittance equivalence and cut set shrinkage.

[0086] Submodule division specifically includes, but is not limited to, the following three categories:

[0087] 1. High-frequency transformer submodule: Considering the admittance relationship between the primary and secondary sides of the transformer and the modeling of the equivalent current source;

[0088] 2. Capacitor (energy storage element) submodule: Discretized using the trapezoidal method and expressed using historical current sources;

[0089] 3. IGBT Device Submodule: The single-port Norton equivalent model is used instead of the device physical model;

[0090] Transformer, capacitor, and IGBT sub-modules are key components of the electromagnetic transient simulation framework. The "simulation framework" itself is a general term for the overall modeling organization and parallel solution mechanism.

[0091] Example 12:

[0092] A method for accurate equivalent modeling and efficient simulation of dual active bridge converters at the device level is presented below:

[0093] First, a high-precision physical model of the IGBT switching device in the DAB converter is established to improve simulation accuracy. The IGBT device model is based on its internal physical characteristics and is equivalently modeled. The IGBT and its anti-parallel diode are considered as a composite device composed of MOSFET and BJT units. The internal junction capacitance, diffusion capacitance, parasitic resistance, and diode recovery characteristics are meticulously modeled to form a high-order nonlinear equivalent circuit that fully reflects the transient behavior of the IGBT during turn-on and turn-off.

[0094] Secondly, addressing the high computational complexity of the high-order IGBT model, a dimensionality reduction and simplification method using cut-set contraction is proposed. This method performs multi-level network simplification on both the IGBT model and the DAB converter circuit. Specifically, the multi-node circuit model of a single IGBT is first equivalent to a Norton equivalent circuit with only one port through cut-set analysis. Then, the overall network formed by embedding multiple equivalent IGBTs into the DAB topology is further simplified through cut-set contraction, significantly reducing the number of network nodes and branch size. Through these two-stage dimensionality reduction processes, the complex device-level model of the DAB converter can be transformed into an equivalent circuit model with equivalent accuracy but a greatly simplified structure.

[0095] Finally, in terms of simulation algorithms, an implicit iterative parallel simulation framework is constructed, which distributes the sub-circuits that have undergone equivalent processing to different computing units for parallel solution, and performs iterative calculations on nonlinear devices at each simulation step to ensure accuracy, thereby greatly improving simulation efficiency.

[0096] Example 13:

[0097] A method for accurate equivalent modeling and efficient simulation of dual active bridge converters at the device level includes the following steps:

[0098] Step 1: Construct a high-precision device-level electromagnetic transient model of IGBT (Insulated Gate Bipolar Transistor) and its anti-parallel diode, and accurately characterize the nanosecond-level switching dynamic behavior of IGBT through implicit iterative algorithm;

[0099] Step 2: Using the cut set contraction dimensionality reduction equivalent method, the high-order nonlinear model of the IGBT is equivalent to a single-port Norton equivalent circuit, that is, a single-port network composed of equivalent admittance and equivalent current source.

[0100] Step 3: Embed the IGBT single-port equivalent circuit into the circuit topology of the dual active bridge (DAB) converter, and perform a second cut-set shrinkage simplification on the embedded DAB circuit to obtain the two-port equivalent circuit model characterizing the DAB converter.

[0101] Step 4: Based on the DAB equivalent circuit model, establish an implicit iterative electromagnetic transient simulation framework and perform simulation solutions in a parallel computing manner to achieve high-speed device-level simulation of the DAB converter.

[0102] The steps for constructing the electromagnetic transient model at the IGBT device level include: based on the physical characteristics of IGBT / FWD (anti-parallel diode), the internal structure of IGBT and its anti-parallel diode is equivalent to a Darlington cascade of MOSFET transistor and BJT transistor; the dynamic characteristics such as junction capacitance, diffusion capacitance and diode tail current are modeled in detail; a high-order nonlinear equivalent circuit containing multiple internal nodes is constructed; and the dynamic response of the IGBT model at each simulation step is obtained by implicit integration and iterative solution.

[0103] The cut-set shrinkage dimensionality reduction equivalent method includes: performing symmetric reconstruction processing on the node admittance matrix obtained after discretization of the IGBT high-order nonlinear equivalent circuit, decomposing the original asymmetric admittance matrix into symmetric components and oblique symmetric components, and correcting the equivalent injected current source according to the oblique symmetric components to obtain a symmetric equivalent admittance matrix and a corrected current source, thereby making the IGBT equivalent to a Norton equivalent circuit with a symmetric node admittance matrix and an accompanying current source.

[0104] After embedding the IGBT single-port equivalent circuit into the DAB converter, the second cut-set shrinkage equivalence of the entire DAB circuit includes: constructing a directed graph reflecting the internal network structure of the DAB converter, selecting one node among the internal nodes as a reference node and at least one node as an external interface node, establishing a cut-set matrix, and shrinking and simplifying the multiple nodes and branches contained inside the DAB converter through node partitioning and circuit equivalence to obtain a two-port equivalent network for connecting the DC input and output terminals of the DAB converter.

[0105] The implicit iterative electromagnetic transient simulation framework includes the following simulation process: First, the high-frequency transformer in the DAB converter is discretized and modeled, and the voltage and current relationship between its primary and secondary sides is equivalent to a decoupled circuit model; then, the capacitor element in the DAB converter is discretized using the trapezoidal integration method to form an equivalent Norton circuit; next, the IGBT device model is equivalently decoupled using the cut-set shrinkage dimensionality reduction method, and each IGBT device is replaced with a single-port Norton equivalent branch with an accompanying current source; then, a second equivalent shrinkage is performed on the DAB circuit containing the IGBT equivalent branch to obtain a simplified DAB equivalent circuit model; finally, the node voltage and branch current of the DAB equivalent circuit model are solved in each simulation step, and the historical state variables of the transformer, capacitor, and IGBT models are updated after the step ends.

[0106] The state update and equivalent calculation processes of each of the equivalent sub-modules (including the transformer sub-module, capacitor sub-module and IGBT device sub-module) are independent of each other and can be executed in parallel, thereby significantly reducing the simulation runtime by utilizing multi-threaded or multi-core processor parallel computing.

[0107] An implicit iterative solution strategy is adopted within each simulation step. The nonlinear characteristics of the IGBT device are repeatedly reduced to equivalent dimensions and the circuit is solved until the preset convergence condition is met, thereby ensuring the accuracy and convergence stability of the simulation results.

[0108] When performing device-level modeling and simulation of a dual active bridge converter containing eight IGBT switching devices across both sides of the full bridge, the multi-layer equivalent dimensionality reduction method can simplify the network, which originally contained dozens of electrical nodes in the detailed model, into an equivalent circuit with only four nodes. This significantly reduces the network size and computational complexity, making the simulation speed comparable to that using an ideal switching model. The number of nodes in the equivalent circuit depends on the topology of the original circuit and the number of external ports to be retained. In typical applications of DAB converters, since the primary and secondary sides are each connected to a pair of DC power supplies or loads, the common scenario after equivalence is to retain four port nodes.

[0109] Example 14:

[0110] A method for accurate equivalent modeling and efficient simulation of dual active bridge converters at the device level, comprising the following:

[0111] Overall Scheme and Topology Analysis of DAB Device-Level Modeling

[0112] like Figure 1 As shown, the DAB converter includes two H-bridge power units: a primary-side full-bridge and a secondary-side full-bridge. Each primary and secondary bridge arm consists of multiple IGBTs and their anti-parallel diodes. A high-frequency transformer connects the primary and secondary sides to achieve bidirectional power transmission and electrical isolation. The DC side also contains energy storage components such as input / output filter capacitors. This invention provides a device-level accurate equivalent modeling and simulation method for the above-mentioned DAB topology, which balances the fine characteristics of switching devices with simulation efficiency in system-level simulations.

[0113] Advanced Modeling and Parameter Extraction of IGBT Devices

[0114] First, a detailed model of the IGBT power switching device is performed at the device level. Figure 2 The equivalent circuit structure of a high-order nonlinear physical model of an IGBT and an anti-parallel diode (FWD) is shown. This model divides the IGBT internally into two parts: a MOSFET unit and a BJT unit. The IGBT's composite transistor structure is represented in Darlington cascade form, including the junction capacitance (e.g., C) that controls the gate-emitter and electrode-collector characteristics. ge C gc The diffusion capacitance C of the BJT transistor diffDynamic components, including gate parasitic resistance, base resistance, and other internal parasitic parameters, are considered. The tail current Irr characteristic of the anti-parallel diode during turn-off is also taken into account, thus enabling a comprehensive characterization of the voltage and current dynamic behavior during IGBT turn-on and turn-off processes. Through methods such as... Figure 3 The parameter extraction process shown can determine the specific parameters of the model components based on the datasheets provided by the device manufacturer and experimental measurement results, enabling the model to quantitatively approximate the characteristics of the actual device. This high-order IGBT equivalent model typically contains eight internal nodes. By writing the KCL / KVL equations of the IGBT's internal circuitry, a set of nonlinear differential equations describing its dynamic processes can be obtained. Because the model introduces rapid transients such as nanosecond-level capacitor charging and discharging and carrier recombination, simulation calculations require a step size much smaller than that of traditional system simulations (reaching tens of nanoseconds or even smaller). Within each step, the aforementioned nonlinear equations are implicitly iterated and solved to capture the detailed transient characteristics of the IGBT.

[0115] Cut set shrinkage equivalent processing of IGBT models

[0116] After obtaining an accurate IGBT device-level model, it needs to be incorporated into the system-level network of the DAB converter for simulation analysis. Directly embedding multiple detailed IGBT models into the system circuit will significantly increase the network size and computational overhead. Therefore, this invention proposes an equivalent dimensionality reduction method based on cut-set contraction, which performs multi-level network simplification on both the IGBT model and the DAB system circuit to minimize simulation complexity while maintaining accuracy. First, equivalent processing is performed on a single IGBT model: Figure 2 The equivalent circuit of the IGBT shown is analyzed at the nodes to obtain the node admittance matrix G (8×8 matrix) and the corresponding associated current source matrix J. Since the IGBT contains both voltage-controlled MOSFET branches and current-controlled BJT branches, the admittance matrix G exhibits asymmetric characteristics, which complicates the circuit solution. This invention uses matrix decomposition to split the asymmetric G into symmetric and oblique-symmetric parts, and accordingly corrects the equivalent current sources, transforming it into a symmetric admittance matrix G. _S And the corrected equivalent current source J'. After this processing, the IGBT model can be represented using a standard discrete equivalent circuit. Next, according to... Figure 4 The diagram shows a directed graph model of the IGBT's internal structure. Nodes ①-⑧ correspond to the eight internal nodes of the IGBT model. Solid lines represent selected branching trees, and dashed lines represent connecting branches. f1 and f7 are the basic cut sets of this directed graph. Based on the connection relationships of this directed graph, cut set equations are written. One node (e.g., node ①) is selected as the external port of the IGBT, and another node (e.g., node ⑧) is selected as the reference node. By eliminating other internal nodes of the IGBT, the equivalent admittance G of the IGBT's external single port can be directly solved._eq and equivalent injected current source I _eq By G _eq and I _eq The Norton equivalent circuit exhibits port characteristics that are completely consistent with the original 8-node IGBT model, but the internal structure is greatly simplified.

[0117] DAB system network dimensionality reduction and equivalent processing

[0118] By embedding the single-port Norton equivalent model of each IGBT into the rectifier bridge and inverter bridge of the DAB converter, the following can be obtained: Figure 5 The DAB converter shown is a primary equivalent circuit. Figure 5 In this circuit, the switching devices on each original bridge arm have been replaced with their Norton equivalent branches. Although the size of the DAB circuit has been reduced through this equivalence, the entire circuit still contains a relatively large number of nodes (e.g., Figure 5 The model shown has approximately 8 nodes. To further improve simulation efficiency, this invention performs a second cut-set shrinkage dimensionality reduction process on the equivalent DAB circuit. Figure 7 The diagram shows a cut-set directed graph after the second-order iso-equivalence of the DAB converter, where black solid lines represent branches of the internal DAB network, red solid lines represent branches acting as external ports, and dashed lines represent connecting branches. According to... Figure 6 The network topology shown establishes cut-set equations. Nodes connected to the external DC power supply or load of the DAB are selected as ports. Further elimination and equivalence are performed on the internal nodes, ultimately transforming the complex internal circuitry of the DAB converter into a two-port network. For example, after the above secondary equivalence processing, the original DAB device-level network model, containing dozens of nodes and consisting of eight IGBT switching devices, can be simplified to a circuit with only four nodes, six branches, and several equivalent current sources. Its two ports correspond to the DC input and DC output terminals of the DAB, respectively. This equivalent model is completely consistent with the complete model in terms of external electrical characteristics, but significantly reduces the node size and connection complexity.

[0119] Design of Parallel Simulation Process Based on Multi-Level Equivalence

[0120] After establishing an equivalent simplified model of the DAB converter, efficient electromagnetic transient simulation calculations can be performed on this basis. To balance accuracy and speed, this invention employs an implicit iterative parallel simulation framework, the overall process of which is as follows: Figure 8As shown. First, based on the selected simulation step size, the high-frequency transformer in the DAB circuit is discretized and modeled. Equivalent admittance and current sources are established based on the voltage and current relationships between the primary and secondary sides of the transformer, thus decoupling the primary and secondary sides during the solution process (corresponding to step 1). Second, the capacitor elements in the DAB topology are discretized using the trapezoidal integral method, and the capacitor branches are represented by equivalent admittance and historical current sources (step 2). Third, the above cut-set dimensionality reduction operation is performed on each IGBT device to obtain its single-port Norton equivalent branch and incorporate it into the network equation (step 3). Subsequently, the entire internal network of the DAB is simplified by a second equivalent reduction, transforming it into a two-port equivalent network model (step 4). After the above preprocessing, the circuit size to be solved at each simulation step is significantly reduced, containing only a very small number of nodes and branches. Next, at each simulation step (corresponding to step 5), the program assembles the node admittance matrix equation containing all equivalent branches and sources, solves for the voltage of each node in the network, and thus calculates the current of each branch. If nonlinear devices such as IGBTs fail to converge in the current step, the simulation program will repeatedly perform equivalent simplification and network solving within the same simulation step, updating the IGBT model state until the convergence condition is met before proceeding to the next long-term calculation. Finally, at the end of each step, the historical variable states of all components are updated as initial values ​​for the next long-term simulation.

[0121] Parallel simulation architecture and computational acceleration

[0122] Because the solutions for each sub-module (such as the high-frequency transformer sub-circuit, each filter capacitor branch, and each IGBT equivalent branch) are independent after the above multi-layer equivalent dimensionality reduction processing, the simulation framework of this invention can fully utilize parallel computing to accelerate the simulation process. The state update and equivalent calculation of each sub-module can be performed simultaneously on different processors or threads. The simulation program only needs to assemble and solve a small-scale aggregate equivalent network at the synchronization point to obtain the system state. This parallel simulation method eliminates the bottleneck of global small-step serial solution and greatly improves computational efficiency. When the DAB model is split into sub-modules, the simulation speedup increases almost linearly with the increase of the number of available computing cores, demonstrating good scalability and engineering practical value.

[0123] Experimental verification and simulation comparison analysis

[0124] The effectiveness of the modeling and simulation method described in this invention has been verified through experimental and simulation comparisons. Figure 9 and Figure 10The voltage and current waveforms of the primary and secondary sides of the high-frequency transformer are compared separately. It can be seen that the maximum deviation between the simulated voltage and current waveforms of the transformer's primary and secondary sides by the model of this invention and the experimentally measured results is no more than 5%, while the simulated waveforms of the traditional ideal switch model have an error of 6% to 10% or more relative to the measured values, significantly higher than the model of this invention. This indicates that the equivalent model of this invention effectively retains the ability of the device-level model to characterize high-frequency transient processes, and the simulation results are closer to the actual behavior. Figure 9 The waveforms of the voltages across the primary and secondary IGBTs are compared. The simulated waveforms of the model in this invention almost match the experimentally measured waveforms, while the traditional ideal switch model shows a significant deviation. This further demonstrates the high accuracy of the model in simulating the transient characteristics of device voltage and current. Figure 10 The figure shows the transient response curve of the DAB converter output voltage when the reference command changes abruptly. It can be seen that during the process of the reference value changing abruptly from 100V to 120V, the output voltage rise slope, transient process and steady-state value simulated by the model of this invention are consistent with the experimental measurement results, which verifies the applicability and accuracy of the model of this invention for control process simulation.

[0125] Simulation efficiency evaluation and engineering value

[0126] Besides its superior accuracy, this invention also demonstrates a significant improvement in simulation efficiency. Under the same time duration, simulating the 1-second operating state of a DAB converter, the method of this invention takes approximately 187.70 seconds, while the traditional PSCAD ideal switching model takes approximately 145.34 seconds—the two are very close. In contrast, directly using an unsimplified device-level physical model for simulation would increase the simulation time by several orders of magnitude under the same conditions, making it almost impossible to complete. Therefore, this invention significantly improves simulation accuracy while achieving a leap in simulation speed, enabling real-time device-level simulation and rapid design evaluation of DAB converters. In summary, the modeling and simulation method proposed in this invention can efficiently and accurately simulate the electromagnetic transient processes of power electronic converters containing numerous high-speed switching devices, possessing broad application prospects and significant engineering value.

Claims

1. A method for accurate equivalent modeling and efficient simulation of a dual active bridge converter at the device level, characterized in that, Includes the following steps: Step 1) Construct the dual active bridge converter circuit topology; Step 2) Construct a high-order nonlinear model of the IGBT in the dual active bridge converter; Step 3) Using the cut set contraction dimensionality reduction equivalent method, the IGBT high-order nonlinear model is equivalent to a single-port Norton equivalent circuit; Step 4) Embed the single-port Norton equivalent circuit into the dual active bridge converter circuit topology, and perform a second cut-set contraction simplification on the embedded dual active bridge converter circuit topology to obtain the dual-port equivalent circuit model characterizing the DAB converter. Step 5) Based on the two-port equivalent circuit model, an implicit iterative electromagnetic transient simulation framework is established. The electromagnetic transient simulation framework is iteratively solved in parallel computing mode to obtain the nanosecond-level switching dynamic response of the IGBT, thereby realizing the device-level high-speed simulation of the DAB converter.

2. The method for accurate equivalent modeling and efficient simulation of a dual active bridge converter at the device level according to claim 1, characterized in that, In step 1), the dual active bridge converter circuit structure includes the H-bridge primary circuit, the high-frequency transformer, and the H-bridge secondary circuit; Both the primary and secondary circuits of the H-bridge include multiple IGBTs; each IGBT has an anti-parallel diode FWD.

3. The method for accurate equivalent modeling and efficient simulation of a dual active bridge converter at the device level according to claim 1, characterized in that, Step 2) involves constructing a high-order nonlinear model of the IGBT, including the following steps: Based on the physical characteristics of IGBT, the internal structure of IGBT is equivalent to a Darlington cascade of MOSFET transistors and BJT transistors, thereby constructing a high-order nonlinear model of IGBT containing multiple internal nodes.

4. The method for accurate equivalent modeling and efficient simulation of a dual active bridge converter at the device level according to claim 1, characterized in that, In step 3), the single-port Norton equivalent circuit includes equivalent admittance and equivalent current source.

5. The method for accurate equivalent modeling and efficient simulation of a dual active bridge converter at the device level according to claim 1, characterized in that, Step 3), which involves converting the high-order nonlinear model of the IGBT into a single-port Norton equivalent circuit, includes the following steps: Step 3.1) Discretize the high-order nonlinear model of the IGBT to obtain the original node admittance matrix; Step 3.2) Decompose the original node admittance matrix into symmetric components and oblique symmetric components, and correct the equivalent injected current source according to the oblique symmetric components to obtain a symmetric equivalent admittance matrix and a corrected current source, thereby transforming the IGBT high-order nonlinear model into a single-port Norton equivalent circuit with a symmetric node admittance matrix and an accompanying current source.

6. The method for accurate equivalent modeling and efficient simulation of dual active bridge converter devices at the S-level according to claim 1, characterized in that, Step 4), the second cut-set contraction and simplification of the embedded dual active bridge converter circuit topology, includes: Step 4.1) Construct a directed graph that reflects the internal network structure of the DAB converter; Step 4.2) Select one internal node as the reference node and at least one internal node as the external interface node to establish a cut set matrix; Step 4.3) By partitioning nodes and performing circuit equivalence operations, the multiple nodes and branches contained within the DAB converter are simplified and reduced to obtain a two-port equivalent circuit model for connecting the DC input and output terminals of the DAB converter.

7. The method for accurate equivalent modeling and efficient simulation of dual active bridge converter devices at the S-level according to claim 6, characterized in that, In step 4.3), when simplifying the multiple nodes and branches contained in the DAB converter, if a node is not directly connected to an external port, control signal, or measurement point, and its adjacent branches can be equivalently expressed by linear or nonlinear elements, then the node is included in the cut set for parallel elimination. Circuit equivalent operations refer to transforming the directed graph of a DAB converter into a standard Norton structure through node rearrangement, matrix splitting, and cut-set transformation, while preserving the admittance G_eq and injected current I_eq of the external ports; G_eq=C t *G*C, I_eq=C t *J; where C is the cut set projection matrix, used to map internal nodes to port nodes; G is the node admittance matrix; and J is the associated current source matrix.

8. The method for accurate equivalent modeling and efficient simulation of a dual active bridge converter at the device level according to claim 1, characterized in that, Step 5), the steps for simulating and solving the electromagnetic transient simulation framework using parallel computing include: The steps for simulating and solving the electromagnetic transient simulation framework using parallel computing include: In each simulation step, the node voltages and branch currents of the two-port equivalent circuit model are solved, and the historical state variables of the transformer, capacitor and IGBT models are updated after the step ends, until the preset convergence conditions are met.

9. The method for accurate equivalent modeling and efficient simulation of a dual active bridge converter at the device level according to claim 8, characterized in that, The convergence condition is that the value of the state variable becomes less than the preset value or the maximum number of iterations is reached.

10. The method for accurate equivalent modeling and efficient simulation of a dual active bridge converter at the device level according to claim 1, characterized in that, The electromagnetic transient simulation framework includes a transformer submodule, a capacitor submodule, and an IGBT device submodule. When performing simulation on the electromagnetic transient simulation framework, the state updates and equivalent calculations of the transformer submodule, capacitor submodule, and IGBT device submodule are performed in parallel.

Images