Multi-port energy router twin model design method based on data driving

Abstract

The invention provides a multi-port energy router twin model design method based on data driving. The method comprises the following steps: constructing a passive device refined model to simulate physical port characteristics; a parameter lumped simulation module is added at a port to simulate a coupling parameter, a parasitic parameter and a line parameter of the whole energy router; obtaining normalized data of port voltage and current of the real object under the corresponding working condition; obtaining normalized data of port voltage and current under corresponding working conditions of the passive device refined model and the parameter lumped simulation module; enabling the normalized data to form a corresponding loss function expression; carrying out minimum value searching on the loss function through a Bayesian optimization algorithm; the steps are repeated, and the parameters under all the working conditions are designed respectively; integrating the parameter data under all working conditions, and taking a mean value of the parameter data as a final parameter design result. The method has high modeling precision, and port characteristics are closer to actual prototype port characteristics.

Description

Technical Field

[0001] The present invention belongs to the technical field of multi-port energy router modeling, and in particular relates to a data-driven multi-port energy router twin model design method. Background Art

[0002] With the rapid development of society and the economy, energy demand continues to increase, and energy crises and environmental issues are becoming increasingly prominent. The widespread application of distributed renewable energy technologies such as photovoltaic power generation, wind power generation, and tidal power generation has made them an important component of renewable energy. However, these distributed renewable energy generation systems are generally characterized by intermittency, randomness, and volatility, and when directly connected to the grid, they may induce bidirectional power flow. This phenomenon poses a severe challenge to traditional power system architectures, which are unable to effectively cope with the complex multi-directional energy flow requirements. Therefore, it is urgent to explore new power system architectures and control strategies to improve adaptability to distributed renewable energy and ensure the stability and reliability of the power system.

[0003] To address these issues, the concept of the Energy Internet emerged, dedicated to building a new energy system. Energy routers, as the core devices of the Energy Internet, are typically composed of power electronic converters and equipped with AC and DC interfaces. This design enables multi-directional energy flow, providing energy isolation, voltage transformation, and plug-and-play functionality. Furthermore, energy routers should be versatile, compatible, and open, supporting energy regulation, scheduling, and distribution at various levels, such as homes, businesses, and campuses. This will meet the needs of local consumption and grid connection of new energy, and promote the integrated scheduling of "source-grid-storage-load." In this way, the Energy Internet will not only improve energy utilization efficiency but also enhance the power system's ability to integrate distributed new energy, providing solid support for achieving sustainable development goals.

[0004] The key device in the Energy Internet, the multi-port energy router, is typically characterized by the use of one or more common DC buses to connect distributed power generation units or energy storage devices via a primary conversion device. This has the advantages of high power density and multi-directional energy flow, while also being able to determine the system's operating status based on the DC bus voltage. Currently, the modeling of the multi-port energy router twin model generally adopts a physical circuit-driven design approach. This physical circuit-driven design effectively simulates the signal transmission and control algorithm execution processes in actual hardware control circuits, restores the characteristics of the DSP controller in the physical prototype, and ensures that the characteristics of the control portion of the constructed twin model are closer to those of the actual circuit control portion.

[0005] However, the physical circuit-driven modeling of the multi-port energy router twin model ignores the influence of stray parameters present in the energy router circuits. In real circuits, as passive components increase in frequency, they generate complex parasitic parameters such as mutual inductance, leakage inductance, crosstalk, transmission line propagation, and reflection. These non-ideal characteristics of components can have undesirable effects on circuit or system performance, significantly affecting the accuracy of the multi-port energy router twin model. Therefore, to construct an accurate energy router twin model, it is necessary to determine the parasitic parameters of each component and circuit in the energy router in order to analyze the characteristics of each port of the energy router. Summary of the Invention

[0006] The present invention discloses a data-driven multi-port energy router twin model design method, which includes the following steps:

[0007] Step 1: Construct a refined model of passive components so that the port characteristics of each module in the refined model are consistent with those of the physical prototype module.

[0008] Step 2: Add a parameter lumped simulation module to the system port of the energy router to simulate the coupling parameters, parasitic parameters, and line parameters of the energy router.

[0009] Step 3: Obtain the port voltage U of the multi-port energy router prototype under the corresponding working conditions r and current I r Normalized data of

[0010] Step 4: The multi-port energy router twin model is constructed by the refined passive device model in step 1 and the whole machine parameter lumping module in step 2, and the port voltage U of the multi-port energy router twin model under the corresponding working condition is collected. s and current I s Normalized data of

[0011] Step 5: Use the normalized data to form the corresponding loss function expression;

[0012] Step 6: Use the Bayesian optimization algorithm to find the minimum value of the loss function;

[0013] Step 7: Repeat the above steps to design the parameters under all working conditions separately;

[0014] Step 8: Integrate the parameter data under all working conditions and take the mean as the final parameter design result.

[0015] Furthermore, in step 1, the passive device refined construction model includes an inductor parasitic parameter model, a capacitor parasitic parameter model and a diode device parasitic parameter model;

[0016] The inductor parasitic parameter model is constructed based on the inductor equivalent model;

[0017] The inductor equivalent model includes the pin parasitic inductance L lead , parasitic capacitance C lead , core equivalent parallel resistance EPR, equivalent parallel capacitance EPC, ideal inductance L dev and the winding equivalent series resistance R s ; Among them, the parasitic capacitance C lead The two ends of the winding are connected in series with the equivalent series resistance R s and the ideal inductor L dev In parallel, in parallel with the core equivalent parallel resistance EPR, in parallel with the equivalent parallel capacitance EPC; and in parallel with the pin parasitic inductance L lead connect;

[0018] In the inductor equivalent model, due to the ideal inductance L dev Much larger than the pin parasitic inductance L lead , ignoring the pin parasitic inductance L lead ; Parasitic capacitance C lead Much smaller than the equivalent parallel capacitance EPC, ignoring the parasitic capacitance C lead ; The inductor parasitic parameter model is simplified to the winding equivalent series resistance R s In series with an ideal inductor;

[0019] Capacitor equivalent model capacitance C Dev In series with the equivalent series resistance ESR and the equivalent series inductance ESL; ignoring the equivalent series inductance ESL, the capacitor parasitic parameter model is simplified to the capacitor C Dev Connected in series with the equivalent series resistance ESR;

[0020] The parasitic parameter model of the diode device is constructed based on the diode equivalent model;

[0021] The diode equivalent model includes the junction capacitance C D , junction resistance R j , body resistance R b and diode D; both ends of diode D are connected to junction resistance R j In parallel with the junction capacitance C D In parallel with the body resistance R b connection; the bulk resistance R b The resistance is very small, the junction resistance R j The resistance is very large and is ignored. The parasitic parameter model of the diode device is simplified to the diode D and the junction capacitance C. D parallel connection.

[0022] Furthermore, in step 2, the parameter lumped simulation module includes the simulation inductor and resistor R L , simulated capacitance and resistance R C , simulated lumped inductance L, simulated lumped capacitance C, 2 input ports and 2 output ports, the 2 input ports are respectively connected to the ports of the energy router, and the 2 output ports are used to output data;

[0023] The first input port is connected to the simulated inductor resistor R L One end, simulates the inductor resistance R L The other end is connected to one end of the simulated lumped inductor L, and the other end of the simulated lumped inductor L is connected to the first output port and the simulated capacitor and resistor R C One end, simulates the capacitance and resistance R C The other end is connected to one end of the simulated lumped capacitor C, and the other end of the simulated lumped capacitor C is connected to the second input port and the second output port respectively.

[0024] Furthermore, in step 5, the loss function is:

[0025] Loss=mse(U r ,U s )+mse(I r ,I s )=f(R L ,L,R C ,C)

[0026] U r is the physical port voltage of the multi-port energy router, I r is the physical port current of the multi-port energy router, U s is the port voltage of the multi-port energy router twin model, I s is the port current of the multi-port energy router twin model, R L is the simulated inductor resistance, L is the simulated lumped inductance value, R C is the simulated capacitance and resistance, and C is the simulated lumped capacitance value; where:

[0027]

[0028] N represents the number of data points, U ri is the voltage of the ith data point of the physical port, U si is the voltage of the i-th data point at the twin model port, I ri is the current of the ith data point of the physical port, I si is the current of the i-th data point of the twin model port.

[0029] Furthermore, in step 6, the Bayesian optimization algorithm specifically proceeds as follows:

[0030] Step 61: Initialize the proxy model and initial dataset;

[0031] Step 62: Use the maximum acquisition function to get the next evaluation point x t ;

[0032] Step 63: Evaluate the objective function value y t =f(x t )+ε t ;

[0033] Step 64: Update dataset D 1:t , and update the probabilistic proxy model;

[0034] Step 65: End.

[0035] Furthermore, in step 62, the Bayesian optimization algorithm maximizes the acquisition function as follows:

[0036] x t =argmax x∈χ α(x|D 1:t-1 )

[0037] Among them, x t represents the next sampling point selected during the t-step optimization, x represents the set of design parameters to be optimized, χ represents the range of design parameters, α(x|D 1:t-1 ) means given existing data D 1:t-1 The value of the acquisition function at the candidate point x under the condition of 1:t-1 Represents the data set that has been collected in t-1 iterations.

[0038] Furthermore, in step 63, the Bayesian optimization algorithm evaluates the objective function expression as follows:

[0039] y t =f(x t )+ε t

[0040] Among them, y t represents the objective function, ε t represents the observation noise.

[0041] Furthermore, in step 64, the Bayesian optimization algorithm data integration updates the proxy model expression as follows:

[0042] D 1:t ={D 1:t-1 ∪(x t ,y t )}

[0043] D 1:t Indicates updating a dataset.

[0044] The beneficial effects achieved by the present invention are:

[0045] This invention addresses the core issues of traditional modeling methods, such as insufficient precision, high complexity, and poor adaptability to operating conditions. Its benefits are directly reflected in improved model accuracy, enhanced parameter calibration efficiency, and enhanced applicability across all scenarios. Port characteristics more closely resemble those of actual prototype ports, providing highly reliable digital twin technology support for energy router applications in smart grids, renewable energy consumption, and other fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 Flowchart of the data-driven multi-port energy router twin model provided in this embodiment;

[0047] Figure 2 Schematic diagram of the equivalent model of the inductance parasitic parameters of the passive device provided in this embodiment

[0048] Figure 3 Schematic diagram of the equivalent model of the capacitance parasitic parameters of the passive device provided in this embodiment;

[0049] Figure 4 Schematic diagram of the equivalent model of parasitic parameters of the passive device diode provided in this embodiment;

[0050] Figure 5 Schematic diagram of the parameter lumped simulation module provided in this embodiment;

[0051] Figure 6 This example provides a block diagram of the multi-module ISOP model of the energy router.

[0052] Figure 7 This is a block diagram of the DCDC module in the multi-module ISOP model of the energy router provided in the example of this embodiment. DETAILED DESCRIPTION

[0053] The present invention will be further described below with reference to specific embodiments, and the advantages and features of the present invention will become clearer as the description proceeds. However, these embodiments are merely exemplary and do not constitute any limitation to the scope of the present invention. It should be understood by those skilled in the art that the details and forms of the technical solutions of the present invention may be modified or replaced without departing from the spirit and scope of the present invention, and such modifications and replacements fall within the scope of protection of the present invention.

[0054] This embodiment provides a data-driven multi-port energy router twin model. This method can quickly and accurately construct an energy router twin model to analyze the characteristics of each port of the energy router. In one embodiment, Figure 6-7As shown in the figure, in the multi-module ISOP solution, the energy router contains five ports. Among them, port 1 is a medium-voltage DC port with a voltage range of 900-4000V, supports bidirectional energy flow, and adopts a half-bridge + CLLC module. The remaining low-voltage ports all use three-phase interleaved parallel Buck modules, of which the power generation port is a unidirectional input port, the energy storage port is a bidirectional port, and the low-voltage DC port and propulsion port are unidirectional output ports. The system bus voltage of this solution is 1000V. The physical circuit drive method is to complete the physical circuit drive by discretizing the internal control of two different sub-modules involved in the multi-module ISOP energy router (simulating DSP control characteristics). These two modules are the half-bridge + CLLC module and the three-phase interleaved parallel Buck module.

[0055] like Figure 1 As shown in the figure, the specific process of the data-driven multi-port energy router twin model is as follows:

[0056] Step 1: Construct a refined model of passive components so that the port characteristics of each module in the energy router simulate the port characteristics of the physical prototype module;

[0057] Furthermore, the passive device refined construction model is as follows:

[0058] Inductor parasitic parameter model:

[0059] The inductor equivalent model that accurately considers parasitic parameters is very complex as an inductor equivalent circuit, as shown below Figure 2 As shown in (a), the inductor equivalent circuit includes the pin parasitic inductance L lead , parasitic capacitance C lead , core equivalent parallel resistance EPR, equivalent parallel capacitance EPC, ideal inductance L dev and the winding equivalent series resistance R s ; Among them, the parasitic capacitance C lead The two ends of the winding are connected in series with the equivalent series resistance R s and the ideal inductor L dev In parallel, in parallel with the core equivalent parallel resistance EPR, in parallel with the equivalent parallel capacitance EPC; and in parallel with the pin parasitic inductance L lead connection; among them, the parasitic parameters are excluding the pin parasitic inductance L lead and parasitic capacitance C lead In addition, there are the core equivalent parallel resistance EPR, the equivalent parallel capacitance EPC and the winding equivalent series resistance R s Since there are many parameters to consider, it is not easy to calculate and measure the parameters, and it is not suitable to model the energy router to analyze the influence of its parasitic parameters on the input current distortion. Therefore, the circuit needs to be further simplified. dev Much larger than the pin parasitic inductance Llead , so the pin parasitic inductance L lead Ignore. Similarly, the parasitic capacitance C lead It is much smaller than the equivalent parallel capacitance EPC, so the parasitic capacitance C lead Also ignored. The typical simplified model of inductor is as follows Figure 2 (b) shown.

[0060] It can be seen that the simplified inductor model is still relatively complex. Considering that the inductor current mainly flows through the winding equivalent series resistance R s , rather than the core equivalent parallel resistance EPR, so the core equivalent parallel resistance EPR can be ignored. And the ideal inductance L dev The operating frequency is much lower than the resonant frequency, and the ideal inductor L dev It will always behave as an inductive resistor, so the equivalent parallel capacitance EPC of the inductor is proportional to the ideal inductance L. dev The influence of can also be ignored. The equivalent model of the final inductance is approximately the winding equivalent series resistance R s With the ideal inductance L dev The series connection, such as Figure 2 (c) shown.

[0061] Capacitor parasitic parameter model

[0062] For input and output capacitors, their typical equivalent models are as follows: Figure 3 As shown, including capacitor C Dev In the capacitor parasitic parameter model, the equivalent series inductance ESL can be ignored and only the capacitor C is included. Dev Connected in series with the equivalent series resistance ESR.

[0063] Diode device parasitic parameter model

[0064] The equivalent model of a typical freewheeling diode is as follows Figure 4 As shown in (a), the diode equivalent model includes the junction capacitance C D , junction resistance R j , body resistance R b and diode D; both ends of the diode are connected to the junction resistance R j In parallel with the junction capacitance C D In parallel with the body resistance R b connection; among them, the parasitic parameters mainly include the junction capacitance C D , junction resistance R j and bulk resistance R b . Body resistance R b The value is within 100mΩ, which is the main reason for the conduction loss of the diode. b The value is very small, the junction resistance R jThe values ​​are very large, so they can be ignored. Finally, the equivalent is the diode D and the junction capacitance C D Parallel connection, see below Figure 4 (b).

[0065] Step 2: Add a parameter lumped simulation module to the system port of the energy router to simulate the coupling parameters, parasitic parameters, line parameters, etc. of the energy router;

[0066] Furthermore, the parameter lumped simulation module is composed of RLC series and parallel connections, and the parasitic parameters, line parameters and coupling parameters in the actual energy router are simulated by adjusting the RLC parameters and the connection method.

[0067] In order to make the lumped RLC parameter simulation module effectively simulate the main characteristics of various parameters and reduce the complexity of the RLC module, this patent performs simplified analysis based on accurate equivalent models of capacitors, inductors, and devices, and constructs a low-order RLC lumped model while maintaining modeling accuracy. The simplified parameter lumped simulation module is shown in the figure below. Figure 5 shown.

[0068] Depend on Figure 5 It can be seen that the parameter lumped simulation module includes the simulation inductor and resistor R L , simulated capacitance and resistance R C , simulated lumped inductance L, simulated lumped capacitance C, 2 input ports and 2 output ports, the 2 input ports are respectively connected to the ports of the energy router, and the 2 output ports are used to output data;

[0069] The first input port is connected to the simulated inductor resistor R L One end, simulates the inductor resistance R L The other end is connected to one end of the simulated lumped inductor L, and the other end of the simulated lumped inductor L is connected to the first output port and the simulated capacitor and resistor R C One end, simulates the capacitance and resistance R C The other end is connected to one end of the simulated lumped capacitor C, and the other end of the simulated lumped capacitor C is connected to the second input port and the second output port respectively.

[0070] Step 3: Obtain the port voltage U of the multi-port energy router prototype (or module) under the corresponding working conditions. r and current I r Normalized data of

[0071] Step 4: The multi-port energy router twin model (or module) constructed by step 1 (passive device refinement model) and step 2 (machine parameter lumping module) is obtained, and the port voltage U of the multi-port energy router twin model (or module) under the corresponding working conditions is collected. s and current I s Normalized data of

[0072] Step 5: Use the normalized data to form the corresponding loss function expression;

[0073] The loss function expression used in this embodiment is as follows:

[0074] Loss=mse(U r ,U s )+mse(I r ,I s )=f(R L ,L,R C ,C)

[0075]

[0076] N represents the number of data points, U ri is the voltage of the ith data point of the physical port, U si is the voltage of the i-th data point at the twin model port, I ri is the current of the ith data point of the physical port, I si is the current of the i-th data point of the twin model port.

[0077] Step 6: Use the Bayesian Optimization algorithm to find the minimum value of the loss function;

[0078] The specific process of the Bayesian optimization algorithm is as follows:

[0079] Step 61: Initialize the proxy model and initial dataset;

[0080] Step 62: Use the maximum acquisition function to get the next evaluation point x t ;

[0081] The expression of the Bayesian optimization algorithm to maximize the acquisition function is as follows:

[0082] x t =argmax x∈χ α(x|D 1:t-1 )

[0083] x t represents the next sampling point selected during the t-step optimization, x represents the set of design parameters to be optimized, χ represents the range of design parameters, α(x|D 1:t-1 ) means given existing data D 1:t-1 The value of the acquisition function at the candidate point x under the condition of 1:t-1 Represents the data set that has been collected in t-1 iterations.

[0084] Step 63: Evaluate the objective function value y t=f(x t )+ε t ;

[0085] y t represents the evaluation objective function value, ε t represents the observation noise.

[0086] The Bayesian optimization algorithm evaluates the objective function expression as follows:

[0087] y t =f(x t )+ε t

[0088] Step 64: Update dataset D 1:t , and update the probabilistic proxy model;

[0089] The expression of the Bayesian optimization algorithm data integration update agent model is as follows:

[0090] D 1:t ={D 1:t-1 ∪(x t ,y t )}

[0091] D 1:t Represents the updated dataset after t iterations.

[0092] Step 65: End.

[0093] Step 7: Repeat the above steps to design the parameters under all working conditions separately;

[0094] Step 8: Integrate the parameter data under all working conditions and take the mean value as the final parameter design result;

[0095] In addition, the terms "upper", "lower", "inner", "outer", "front", and "back" are used for descriptive purposes only and should not be understood as indicating or implying relative importance. Unless otherwise specifically stated, the relative steps, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of the present invention.

[0096] Of course, the above description is only a specific embodiment of the present invention and is not intended to limit the scope of implementation of the present invention. Any equivalent changes or modifications made based on the structure, features and principles described in the scope of the patent application of the present invention should be included in the scope of the patent application of the present invention.

[0097] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the above-described embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily conceive of changes to the technical solutions described in the above-described embodiments within the technical scope disclosed by the present invention, or replace some of the technical features therein with equivalents. Such modifications, changes, or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A data-driven multi-port energy router twin model design method, characterized in that: The data-driven multi-port energy router twin model design method includes the following steps: Step 1: Construct a refined model of passive components so that the port characteristics of each module in the refined model are consistent with those of the physical prototype module. Step 2: Add a parameter lumped simulation module to the system port of the energy router to simulate the coupling parameters, parasitic parameters, and line parameters of the energy router. Step 3: Obtain the port voltage U of the multi-port energy router prototype under the corresponding working conditions r and current I r Normalized data of Step 4: The multi-port energy router twin model is constructed by the refined passive device model in step 1 and the whole machine parameter lumping module in step 2, and the port voltage U of the multi-port energy router twin model under the corresponding working condition is collected. s and current I s Normalized data of Step 5: Use the normalized data to form the corresponding loss function expression; Step 6: Use the Bayesian optimization algorithm to find the minimum value of the loss function; Step 7: Repeat the above steps to design the parameters under all working conditions separately; Step 8: Integrate the parameter data under all working conditions and take the mean as the final parameter design result.

2. The data-driven multi-port energy router twin model design method according to claim 1 is characterized in that: In step 1, the passive device refined construction model includes an inductor parasitic parameter model, a capacitor parasitic parameter model and a diode device parasitic parameter model; The inductor parasitic parameter model is constructed based on the inductor equivalent model; The inductor equivalent model includes the pin parasitic inductance L lead , parasitic capacitance C lead , core equivalent parallel resistance EPR, equivalent parallel capacitance EPC, ideal inductance L dev and the winding equivalent series resistance R s ; Among them, the parasitic capacitance C lead The two ends of the winding are connected in series with the equivalent series resistance R s and the ideal inductor L dev In parallel, in parallel with the core equivalent parallel resistance EPR, in parallel with the equivalent parallel capacitance EPC; and in parallel with the pin parasitic inductance L lead connect; In the inductor equivalent model, due to the ideal inductance L dev Much larger than the pin parasitic inductance L lead , ignoring the pin parasitic inductance L lead ; Parasitic capacitance C lead Much smaller than the equivalent parallel capacitance EPC, ignoring the parasitic capacitance C lead ; The inductor parasitic parameter model is simplified to the winding equivalent series resistance R s In series with an ideal inductor; Capacitor equivalent model capacitance C Dev In series with the equivalent series resistance ESR and the equivalent series inductance ESL; ignoring the equivalent series inductance ESL, the capacitor parasitic parameter model is simplified to the capacitor C Dev Connected in series with the equivalent series resistance ESR; The parasitic parameter model of the diode device is constructed based on the diode equivalent model; The diode equivalent model includes the junction capacitance C D , junction resistance R j , body resistance R b and diode D; both ends of diode D are connected to junction resistance R j In parallel with the junction capacitance C D In parallel with the body resistance R b connection; the bulk resistance R b The resistance is very small, the junction resistance R j The resistance is very large and is ignored. The parasitic parameter model of the diode device is simplified to the diode D and the junction capacitance C. D parallel connection.

3. The data-driven multi-port energy router twin model design method according to claim 1 is characterized in that: In step 2, the parameter lumped simulation module includes the simulated inductor and resistor R L , simulated capacitance and resistance R C , simulated lumped inductance L, simulated lumped capacitance C, 2 input ports and 2 output ports, the 2 input ports are respectively connected to the ports of the energy router, and the 2 output ports are used to output data; The first input port is connected to the simulated inductor resistor R L One end, simulates the inductor resistance R L The other end is connected to one end of the simulated lumped inductor L, and the other end of the simulated lumped inductor L is connected to the first output port and the simulated capacitor and resistor R C One end, simulates the capacitance and resistance R C The other end is connected to one end of the simulated lumped capacitor C, and the other end of the simulated lumped capacitor C is connected to the second input port and the second output port respectively.

4. The data-driven multi-port energy router twin model design method according to claim 1 is characterized in that: In step 5, the loss function is: Loss=mse(U r ,U s )+mse(I r ,I s )=f(R L ,L,R C ,C) U r is the physical port voltage of the multi-port energy router, I r is the physical port current of the multi-port energy router, U s is the port voltage of the multi-port energy router twin model, I s is the port current of the multi-port energy router twin model, R L is the simulated inductor resistance, L is the simulated lumped inductance value, R C is the simulated capacitance and resistance, and C is the simulated lumped capacitance value; where: N represents the number of data points, U ri is the voltage of the ith data point of the physical port, U si is the voltage of the i-th data point at the twin model port, I ri is the current of the ith data point of the physical port, I si is the current of the i-th data point of the twin model port.

5. The data-driven multi-port energy router twin model design method according to claim 1 is characterized in that: In step 6, the specific process of the Bayesian optimization algorithm is as follows: Step 61: Initialize the proxy model and initial dataset; Step 62: Use the maximum acquisition function to get the next evaluation point x t ; Step 63: Evaluate the objective function value y t =f(x t )+ε t ; Step 64: Update dataset D 1:t , and update the probabilistic proxy model; Step 65: End.

6. The data-driven multi-port energy router twin model design method according to claim 5 is characterized in that: In step 62, the expression of the acquisition function maximized by the Bayesian optimization algorithm is as follows: x t =argmax x∈χ α(x|D 1:t-1 ) Among them, x t represents the next sampling point selected during the t-step optimization, x represents the set of design parameters to be optimized, χ represents the range of design parameters, α(x|D 1:t-1 ) means given existing data D 1:t-1 The value of the acquisition function at the candidate point x under the condition of 1:t-1 Represents the data set that has been collected in t-1 iterations.

7. The data-driven multi-port energy router twin model design method according to claim 6 is characterized in that: In step 63, the Bayesian optimization algorithm evaluates the objective function expression as follows: y t =f(x t )+ε t Among them, y t represents the objective function, ε t represents the observation noise.

8. The data-driven multi-port energy router twin model design method according to claim 7 is characterized in that: In step 64, the Bayesian optimization algorithm data integration update agent model expression is as follows: D 1:t ={D 1:t-1 ∪(x t ,y t )} D 1:t Indicates updating a dataset.

Images