High-efficiency parameter identification method for dual-active bridge converter based on sparse state acquisition and fast state deduction

Abstract

The invention relates to the technical field of dual active bridge converter parameter identification, in particular to a dual active bridge converter efficient parameter identification method based on sparse state acquisition and fast state deduction, which comprises the following steps: S101, defining model input; under the conventional sampling configuration of the MCU, the DAB converter collects a system state when a modulation wave and a carrier wave converge each time; s102, carrying out dynamic modeling on the DAB converter; s103, updating the time step of the system state; discretization processing is carried out to obtain a system state evolution relation; s104, performing population initialization; s105, carrying out population target fitness evaluation; s106, carrying out genetic evolution; comprising the steps of pairing selection, crossover and variation, and generating a progeny population with the scale of M; s107, judging a termination condition and outputting an optimal solution; when the fitness of the individual with the optimal target performance tends to converge, the algorithm is terminated; if not, returning to S105 to continue loop iteration; s108, extracting parameters; and after algorithm convergence, the parameter representing the optimal individual on the main target is a final parameter identification result.

Description

Technical Field

[0001] This invention relates to the field of parameter identification technology for dual active bridge converters, and more specifically, to an efficient parameter identification method for dual active bridge converters based on sparse state acquisition and fast state deduction. Background Technology

[0002] Dual Active Bridge (DAB) DC-DC converters, with their advantages of electrical isolation, bidirectional power flow capability, high power density, and ease of soft switching, have become core power conversion devices in critical applications such as on-board chargers for electric vehicles, large-scale energy storage systems, solid-state transformers, and DC microgrids. Under long-term, high-frequency, high-voltage, and complex operating conditions, the converter's key physical parameters (such as phase-shifting inductance, high-frequency transformer leakage inductance, output filter capacitor value and its equivalent series resistance (ESR), and secondary-side load resistance) undergo significant dynamic drift due to the combined effects of magnetic component thermal effects, dielectric aging, and ambient temperature drift. The accuracy of these parameters is not only a prerequisite for implementing high-performance control strategies such as Model Predictive Control (MPC), but also the foundation for converter state of health (SOH) monitoring and fault early warning. Therefore, constructing a high-fidelity system model and achieving accurate online identification of multiple parameters is of paramount importance for ensuring the safe and efficient operation of DAB converters.

[0003] However, existing parameter identification techniques still face significant challenges in data acquisition and computation methods. Regarding data acquisition, DAB converters, as high-order nonlinear systems, contain abundant high-frequency harmonic components in their voltage and current waveforms. To obtain sufficient information to invert system parameters, traditional methods typically rely on high-sampling-rate analog-to-digital converters (ADCs) or dedicated high-frequency observation equipment to capture complete waveform details. However, in industrial embedded control systems (such as DSP- or MCU-based controllers), the sampling frequency is often insufficient to meet the requirements of high-frequency reconstruction due to limitations in hardware cost and chip processing power.

[0004] In terms of computational algorithms, existing solutions for multi-parameter identification problems are mainly divided into two categories: analytical derivation methods and intelligent optimization algorithms, but both have limitations. While analytical methods are fast, they rely excessively on idealized mathematical models and are poorly robust to non-ideal factors (such as parasitic parameters). Intelligent evolutionary algorithms, such as Genetic Algorithms (GA) and Particle Swarm Optimization (PSO), while possessing global optimization capabilities and able to handle nonlinear problems, have a high computational load. Simply using numerical integration methods for forward derivation can accurately simulate dynamic processes, but lacks an automatic optimization mechanism; relying solely on evolutionary algorithms can easily lead to local optima.

[0005] In summary, the key bottleneck to achieving low-cost, high-precision parameter identification of DAB converters lies in how to extract deep physical characteristics of a system under extremely sparse data constraints. Existing identification methods have not yet found a balance between cost and accuracy, or speed and precision, making large-scale deployment in cost-sensitive industrial products difficult. Therefore, how to achieve rapid and accurate identification of multiple parameters of DAB converters using sparse sampling data without changing existing conventional sampling configurations or introducing additional high-frequency detection hardware is a pressing technical challenge that needs to be overcome, and it also has significant engineering application value and industrialization prospects. Summary of the Invention

[0006] This invention addresses the problem that existing parameter identification methods for dual active bridge converters rely on high-bandwidth sampling equipment to obtain dense data, and struggle to balance computational speed and identification accuracy when dealing with nonlinear, strongly coupled systems. This invention proposes an efficient parameter identification method for dual active bridge converters based on sparse state acquisition and fast state deduction.

[0007] The objective of this invention is achieved through the following technical solution:

[0008] An efficient parameter identification method for dual active bridge converters based on sparse state acquisition and fast state deduction includes the following steps:

[0009] S101, Define model input; Under the MCU's normal sampling configuration, the DAB converter collects the system state when each switch is turned on;

[0010] S102, Dynamic Modeling of DAB Converter; Constructing a Continuous-Time State-Space Model of the DAB Converter;

[0011] S103, Time step update of system state; The above model is discretized using the Runge-Kutta method to obtain the system state evolution relationship;

[0012] S104, Population initialization; The initial population consists of 2*M individuals, each of which is a complete set of parameter vectors for the model to be identified;

[0013] S105, Population target fitness assessment; Construct a single-target fitness function, which is calculated based on the sum of squared errors between the model calculation values ​​derived in step S103 and the measured sample values ​​obtained in step S101;

[0014] S106, genetic evolution; including pairing selection, crossover, and mutation, generating a progeny population of size M;

[0015] S107, Termination condition judgment and optimal solution output; The algorithm terminates when the fitness of the individual with the best performance in the target converges; If it has not yet converged, it returns to S105 to continue the loop iteration.

[0016] S108, Extract parameters; After the algorithm converges, select the individual with the best performance on the main target, and its parameters are the final parameter identification results.

[0017] Furthermore, in step S101, the collected system state parameters include the inductor current i L Secondary side output voltage V2, switching sequence signals S1, S2, time point t and time interval Δt for each switching state.

[0018] Furthermore, in step S102, the continuous-time state-space model of the DAB converter is represented as follows:

[0019]

[0020] Wherein, the inductor current i L The secondary-side output voltage V2 is the system state variable of the DAB converter; the parameters to be identified are the inductor L, the secondary-side supporting capacitor C, and the parasitic resistance R of the supporting capacitor. C 1. Load resistance R; V1 is the DC power supply, and K is the transformer turns ratio.

[0021] Furthermore, the specific steps of S103 are as follows:

[0022] First, at each time step Calculate i separately L The intermediate slope of the q-th order of V2:

[0023]

[0024]

[0025] Secondly, the next time step t is calculated using the calculated slopes of the q-group. n+1 System status values:

[0026]

[0027]

[0028] Among them, {a 11 ,…,a qq}, {b1,…,b q}, {c1,…,c q The value of} can be found in the RK table.

[0029] Furthermore, in step S104, the parameter vector of the model to be identified is P=[L,R]c [C,R], the individuals in the initial population are obtained by uniform random sampling in a logarithmic coordinate system.

[0030] Furthermore, the method used for uniform random sampling in logarithmic coordinates for the individuals in the initial population is as follows:

[0031] First, physical search boundaries for each parameter are set according to the design specifications of the DAB converter. These physical search boundaries include an upper bound UB and a lower bound LB. Then, for each parameter p to be generated... j In its logarithmic interval [log 10 (LB j ),log 10 (UB j Generate a uniformly distributed random number within the []; finally, map the random number back to the original parameter value through antilogarithmic operation.

[0032] Furthermore, in step S105, the single-objective fitness function is:

[0033]

[0034] in,{ , } represents the model calculation value obtained in step S103, { , } represents the measured sampled value obtained in step S101, N is the number of switching cycles, 4N+1 is the total number of sampling points, w is the weighting coefficient for the difference in magnitude between the voltage and current, and the fitness value J(P) is... i The smaller the value, the more outstanding the individual.

[0035] Furthermore, in step S106, the method for pairing and selecting is as follows:

[0036] Two individuals are randomly and repeatedly selected from a parent population of size M; the two individuals are compared, and the one with the smaller fitness value J(Pi) is judged to be the winner; the winner is placed into a pairing pool; this process is repeated M times to form a pairing pool of size M.

[0037] The parent population is the initial population or the offspring population generated in the previous iteration.

[0038] Furthermore, in step S106, the crossover operation is performed by performing the crossover operation on each pair of parent individuals P1=[p 1,1 ,...,p 1,j ] and P2=[p 2,1 ,...,p 2,j ], simulating binary interleaving operations;

[0039] The method for simulating binary crossover is as follows:

[0040] Generate a random number u between [0, 1];

[0041] Based on a preset distribution index η c Calculate an expansion factor β according to the following formula:

[0042] (7)

[0043] For each parameter p in the parent vector j Two new offspring parameters c are generated according to the following formula. 1,j and c 2,j :

[0044] (8)

[0045] (9).

[0046] Furthermore, in step S106, the crossover operation is performed using a polynomial mutation operator compatible with the simulated binary crossover operation; for each offspring individual C generated after crossover, a small mutation probability p is applied. m Decide whether to mutate it; if so, mutate it for each parameter c in the individual. j Independent operation, the operation steps are as follows:

[0047] Generate a random number r between [0, 1].

[0048] Based on a preset variation distribution index η m Calculate a disturbance δ according to the following formula;

[0049] (10)

[0050] The new parameter c after mutation is generated according to the following formula. ’ j :

[0051] (11)

[0052] Among them, UB j and LB j It is parameter p j The upper and lower bounds of the search, if c ’ j If it exceeds the limit, set it as the boundary value.

[0053] Compared with the prior art, the beneficial effects of the present invention are:

[0054] The method in this invention, without introducing any additional high-frequency monitoring hardware, utilizes a sparse sampling mechanism combined with a deep fusion strategy of high-order Runge-Kutta algorithm and Genetic Algorithm (GA) to achieve accurate monitoring of key physical parameters of the DAB converter (including inductor L, secondary-side output resistance R, output capacitance C and its parasitic resistance R). c ( ) fast and accurate identification.

[0055] This invention overcomes the limitations of high-frequency sampling, achieving low-cost sensing based on sparse data. It proposes a "sparse state acquisition" mechanism, requiring only the acquisition of instantaneous values ​​of inductor current and output voltage at specific moments during the switching action of the DAB converter. By deriving the discrete-time state evolution law based on the switching sequence, this method can extract key information characterizing the global dynamic characteristics of the system from extremely sparse sampling points. This method is fully compatible with the conventional sampling configuration of existing industrial-grade MCUs, requiring no modification to the hardware topology or the addition of auxiliary circuits, significantly reducing the deployment threshold and hardware cost of the algorithm, and possessing extremely high engineering application value.

[0056] This invention integrates numerical deduction and evolutionary optimization, achieving a dual improvement in computational efficiency and identification accuracy. The invention proposes a strategy of "integrating rapid state deduction and evolutionary optimization." On one hand, a high-fidelity state deduction engine is constructed using the high-order Runge-Kutta method. This engine can reconstruct the nonlinear dynamic trajectory of the system, including the influence of parasitic parameters, with large step sizes and high precision among sparse sampling points, solving the problem of information loss under sparse data and ensuring the physical fidelity of parameter estimation. On the other hand, this deduction engine is embedded into the fitness evaluation stage of a genetic algorithm (GA), utilizing the global search capability of GA to overcome the local extremum problem caused by multi-parameter coupling. This organic integration of "numerical deduction + evolutionary computation" accelerates the calculation process of a single state update by utilizing the high-order truncation error characteristics of RK, and ensures global convergence using GA, thereby achieving efficient and accurate identification of complex parameters of the DAB converter with limited computational resources. Attached Figure Description

[0057] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0058] Figure 1 This is a flowchart of the efficient parameter identification method for dual active bridge converters based on sparse state acquisition and fast state deduction in this invention.

[0059] Figure 2 This is a schematic diagram of the dynamic process and sampling data of the DAB converter in this invention.

[0060] Figure 3 This is the DAB converter topology diagram in this invention.

[0061] Figure 4 This is a schematic diagram of the data loop iteration process in this invention. Detailed Implementation

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] Example:

[0064] like Figure 1 As shown, this invention proposes an efficient parameter identification method for dual active bridge converters based on sparse state acquisition and fast state deduction, comprising the following steps:

[0065] S101, Define the model input:

[0066] like Figure 2 As shown, under the normal sampling configuration of the MCU, the DAB converter will collect the system status, including the inductor current i, when each switch is turned on. L Secondary side output voltage V2, switching sequence signals S1, S2, time point t and time interval Δt for each switching state.

[0067] S102, Dynamic Modeling of DAB Converter:

[0068] like Figure 3 Construct a continuous-time state-space model of the DAB converter and represent it as follows:

[0069]

[0070] Wherein, the inductor current i L The secondary-side output voltage V2 is the system state variable of the DAB converter; the parameters to be identified are the inductor L, the secondary-side supporting capacitor C, and the parasitic resistance R of the supporting capacitor. C Load resistance R. S1 is the switching status signal of the primary-side controllable switches (Q1, Q2, Q3, Q4). When Q1 and Q4 are on, S1=1; when Q2 and Q3 are on, S1=-1. S2 is the switching status signal of the secondary-side controllable switches (Q5, Q6, Q7, Q8). When Q5 and Q8 are on, S2=1; when Q6 and Q7 are on, S2=-1. V1 is the DC power supply, and K is the transformer turns ratio.

[0071] S103, System status time step update:

[0072] like Figure 4 Furthermore, the above model is discretized using the q-order Runge-Kutta (RK) method to obtain the system state evolution relationship for each switching state. First, i is calculated in each time step. L The intermediate slope of the q-th order of V2:

[0073]

[0074]

[0075] Secondly, the next time step t is calculated using the calculated slopes of the q-group. n+1 System status values:

[0076]

[0077]

[0078] Among them, {a 11 ,…,a qq}, {b1,…,b q}, {c1,…,c q The value of} can be found in the RK table.

[0079] S104, Population Initialization:

[0080] The population consists of 2*M individuals, each of which is a complete set of parameter vectors P=[L,R] for the model to be identified. c [C,R]. The physical values ​​of the parameters of the DAB converter span multiple orders of magnitude. This invention employs a method of uniform random sampling in a logarithmic coordinate system.

[0081] In practice, the physical search boundaries for each parameter are first set according to the design specifications of the DAB converter. The value criteria in this embodiment are as follows:

[0082] Inductance L: Set range [1uH, 1mH] (i.e., LB1 = 10) -6 UB1=10 -3 ), covering the range of common high-frequency leakage inductance;

[0083] Parasitic resistance R c : Set range [1mΩ, 1Ω] (i.e., LB2 = 10) -4 ,UB2=1), set according to the typical ESR characteristics of film capacitors or electrolytic capacitors;

[0084] Capacitor C: Set range [1uF, 1mF] (i.e., LB3=10) -5 UB3=10 -2 (Set according to output voltage ripple requirements);

[0085] Load resistance R: Set range [1Ω, 1000Ω] (i.e., LB4=1, UB4=10) 3 (), covering operating conditions from heavy load to light load.

[0086] Then, for each parameter p to be generated j (j=1,2,3,4), in its logarithmic interval [log 10 (LB j ),log 10 (UB j A uniformly distributed random number is generated within the []]. Finally, this random number is mapped back to the original parameter value through an antilogarithmic operation (i.e., a power of 10), thus completing the construction of the initial population.

[0087] S105, Population Target Fitness Assessment:

[0088] To identify the optimal parameter vector P, a single-objective fitness function J(P) is constructed. i This function is based on the model calculation values ​​obtained in step S103 using the Runge-Kutta (RK) method. , } and the measured sampled values ​​obtained in step S101 { , The sum of squared errors between} is calculated:

[0089]

[0090] Where N is the number of switching cycles, and 4N+1 is the total number of sampling points. w is a weighting coefficient for the difference in magnitude between voltage and current, aiming to normalize the error terms of voltage and current to the same order of magnitude. In practice, it is recommended to use the reciprocal of the square of the ratio of rated voltage to rated current. Fitness value J(P) i The smaller the value, the more outstanding the individual.

[0091] S106, genetic evolution:

[0092] This step aims to inherit the superior genes (parameter combinations) from the parent generation and explore new regions of the parameter space by introducing random perturbations, thereby preventing the algorithm from prematurely converging to local optima. Genetic evolution operations mainly include three core steps: pairing selection, crossover, and mutation.

[0093] Pairing Selection: Before generating offspring, individuals need to be selected from a parent population of size M for pairing and reproduction. This invention employs a binary tournament selection strategy. This strategy repeats the following three steps M times to build a pairing pool of size M:

[0094] 1. Randomly and repeatedly select two individuals from the parent population;

[0095] 2. Compare the merits and demerits of these two individuals, and their fitness values ​​J(P) i The smaller one is the winner;

[0096] 3. Place the winning individuals into the pairing pool.

[0097] Crossover: The crossover operation simulates the exchange of chromosomes during biological reproduction, aiming to combine the superior traits of two parent individuals to produce potentially better offspring. This invention employs a simulated binary crossover operator optimized for real numbers. For each pair of parent individuals in the pairing pool, P1 = [p 1,1 ,...,p 1,j ] and P2=[p 2,1 ,...,p 2,j The simulated binary crossover operation is as follows:

[0098] 1. Generate a random number u between [0, 1];

[0099] 2. Preset a distribution index η c This index controls the distance between offspring and parents, η c The larger the value, the closer the offspring are to the parent. Based on the general standards of real-number encoded genetic algorithms, a typical value of 1 achieves the best balance between preserving the parent's superior patterns and exploring new solutions. According to this preset distribution index η... c Calculate an expansion factor β:

[0100] (7)

[0101] 3. For each parameter p in the parent vector j (j=1,2,3,4), generate two new offspring parameters c 1,j and c 2,j :

[0102] (8)

[0103] (9)

[0104] Mutation: Mutation simulates gene mutation by applying a small random perturbation to the parameters of offspring individuals, introducing new genes into the population. It is crucial for maintaining population diversity and escaping local optima. This invention employs a polynomial mutation operator compatible with the simulated binary crossover operation. For each offspring individual C generated after crossover, a small mutation probability p is applied. m (This embodiment sets a typical value p) m =0.1) determines whether to mutate it. According to the theory of genetic algorithms, the mutation probability should be such that on average, one gene of each individual is mutated. Too high a probability will cause the algorithm to degenerate into a random search, while too low a probability will easily get stuck in a local optimum.

[0105] If mutation is performed, then for each parameter c in the individual j Independent operation:

[0106] 1. Generate a random number r between [0, 1];

[0107] 2. Based on a preset variation distribution index η m (This embodiment sets a typical value η) m =1), the principle is the same as the cross operation, used to control the amplitude of the variation disturbance to be concentrated near the original value, and to calculate a disturbance amount δ;

[0108] (10)

[0109] 3. Generate the new parameter c after mutation. ’ j Among them, UB j and LB j It is parameter p j The upper and lower bounds of the search for (j=1,2,3,4). If c ’ j If it exceeds the boundary, set it as the boundary value:

[0110] (11)

[0111] By performing the above genetic evolution operations, a completely new offspring population of size M is finally generated.

[0112] S107, Termination condition determination and optimal solution output:

[0113] This step monitors the evolutionary process and terminates the algorithm when specific conditions are met. The fitness J(P) of the individual that best performs in the target region is determined by this process. i The algorithm tends to converge, meaning it terminates when the fitness difference after Q consecutive iterations is less than a set threshold ε. In this embodiment, Q=20 and ε=10. -6This threshold is set based on the accuracy requirements of the fitness function to ensure that the parameter identification results remain stable within the range of significant numbers. If convergence has not yet occurred, return to S105 to continue the iterative loop.

[0114] S108: Extract parameters

[0115] After the algorithm converges, the individual with the best performance on the main target is selected, and its parameters are the final parameter identification results. The parameter identification results in this embodiment are shown in the table below.

[0116]

[0117] The relative identification errors for all four parameters are less than 2%, with the errors for inductance, capacitance, and load resistance even within 0.3%. This demonstrates that the proposed method can achieve extremely high parameter identification accuracy even under sparse data conditions.

[0118] The method in this invention features extremely low hardware cost and excellent compatibility. It eliminates the need for high-frequency ADCs or additional detection circuits, directly utilizing the sparse switching signals of existing MCUs. It only requires acquiring the instantaneous values ​​of inductor current and output voltage at specific moments during the switching action of the DAB converter. In this embodiment, only 5 samples are needed per switching cycle, facilitating engineering deployment. It combines computational efficiency and accuracy, using physical deduction to compensate for data sparsity and combining evolutionary algorithms to achieve global optimization, thus resolving the contradiction between speed and accuracy. Deep decoupling of multiple parameters accurately separates inductance, capacitance, and key parasitic resistance, significantly improving the model's ability to capture transient details. It supports online health monitoring, effectively sensing the aging status of devices by tracking parameter drift in real time, thereby improving reliability throughout the entire lifecycle.

[0119] This invention breaks through the dependence of traditional DAB converter identification methods on high-frequency sampling, and creatively integrates the Runge-Kutta method and genetic algorithm to achieve "fast and accurate" identification by using only sparse data at MCU switching moments.

[0120] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for efficient parameter identification of a dual active bridge converter based on sparse state acquisition and fast state deduction, characterized in that, Includes the following steps: S101, Define model input; Under the MCU's normal sampling configuration, the DAB converter collects the system state when each switch is turned on; S102, Dynamic Modeling of DAB Converter; Constructing a Continuous-Time State-Space Model of the DAB Converter; S103, Time step update of system state; The above model is discretized using the Runge-Kutta method to obtain the system state evolution relationship; S104, Population initialization; The initial population consists of 2*M individuals, each of which is a complete set of parameter vectors for the model to be identified; S105, Population target fitness assessment; Construct a single-target fitness function, which is calculated based on the sum of squared errors between the model calculation values ​​derived in step S103 and the measured sample values ​​obtained in step S101; S106, genetic evolution; This includes pairing selection, crossover, and mutation, generating a progeny population of size M; S107, Termination condition judgment and optimal solution output; The algorithm terminates when the fitness of the individual with the best performance in the target converges; If it has not yet converged, it returns to S105 to continue the loop iteration. S108, Extract parameters; After the algorithm converges, select the individual with the best performance on the main target, and its parameters are the final parameter identification results.

2. The efficient parameter identification method for dual active bridge converters based on sparse state acquisition and fast state deduction as described in claim 1, characterized in that, In step S101, the collected system state parameters include the inductor current i L Secondary side output voltage V2, switching sequence signals S1, S2, time point t and time interval Δt for each switching state.

3. The efficient parameter identification method for dual active bridge converters based on sparse state acquisition and fast state deduction as described in claim 2, characterized in that, In step S102, the continuous-time state-space model of the DAB converter is represented as follows: Wherein, the inductor current i L The secondary-side output voltage V2 is the system state variable of the DAB converter; the parameters to be identified are the inductor L, the secondary-side supporting capacitor C, and the parasitic resistance R of the supporting capacitor. C 1. Load resistance R; V1 is the DC power supply, and K is the transformer turns ratio.

4. The efficient parameter identification method for dual active bridge converters based on sparse state acquisition and fast state deduction according to claim 3, characterized in that, The specific steps of S103 are as follows: First, at each time step Calculate i separately L The intermediate slope of the q-th order of V2: Secondly, the next time step t is calculated using the calculated slopes of the q-group. n+1 System status values: Among them, {a 11 ,…,a qq }, {b1,…,b q }, {c1,…,c q The value of} can be found in the RK table.

5. The efficient parameter identification method for dual active bridge converters based on sparse state acquisition and fast state deduction according to claim 4, characterized in that, In step S104, the parameter vector of the model to be identified is P=[L,R] c [C,R], the individuals in the initial population are obtained by uniform random sampling in a logarithmic coordinate system.

6. The efficient parameter identification method for dual active bridge converters based on sparse state acquisition and fast state deduction according to claim 5, characterized in that, The method used for uniform random sampling of individuals in the initial population in a logarithmic coordinate system is as follows: First, physical search boundaries for each parameter are set according to the design specifications of the DAB converter. These physical search boundaries include an upper bound UB and a lower bound LB. Then, for each parameter p to be generated... j In its logarithmic interval [log 10 (LB j ),log 10 (UB j Generate a uniformly distributed random number within the []; finally, map the random number back to the original parameter value through antilogarithmic operation.

7. The efficient parameter identification method for dual active bridge converters based on sparse state acquisition and fast state deduction according to claim 6, characterized in that, In step S105, the single-target fitness function is: in,{ , } represents the model calculation value obtained in step S103, { , } represents the measured sampled value obtained in step S101, N is the number of switching cycles, 4N+1 is the total number of sampling points, w is the weighting coefficient for the difference in magnitude between the voltage and current, and the fitness value J(P) is... i The smaller the value, the more outstanding the individual.

8. The efficient parameter identification method for dual active bridge converters based on sparse state acquisition and fast state deduction according to claim 7, characterized in that, In step S106, the method for pairing and selecting is as follows: Two individuals are randomly and repeatedly selected from a parent population of size M; the two individuals are compared, and the one with the smaller fitness value J(Pi) is judged to be the winner; the winner is placed into a pairing pool; this process is repeated M times to form a pairing pool of size M. The parent population is the initial population or the offspring population generated in the previous iteration.

9. The efficient parameter identification method for dual active bridge converters based on sparse state acquisition and fast state deduction according to claim 8, characterized in that, In step S106, the crossover operation is performed by performing the crossover operation on each pair of parent individuals P1=[p 1,1 ,...,p 1,j ] and P2=[p 2,1 ,...,p 2,j ], simulating binary interleaving operations; The method for simulating binary crossover is as follows: Generate a random number u between [0, 1]; Based on a preset distribution index η c Calculate an expansion factor β according to the following formula: （7） For each parameter p in the parent vector j Two new offspring parameters c are generated according to the following formula. 1,j and c 2,j : （8） （9）。 10. The efficient parameter identification method for dual active bridge converters based on sparse state acquisition and fast state deduction according to claim 9, characterized in that, In step S106, the crossover operation is performed using a polynomial mutation operator compatible with the simulated binary crossover operation; for each offspring C generated after crossover, a small mutation probability p is applied. m Decide whether to mutate it; if so, mutate it for each parameter c in the individual. j Independent operation, the operation steps are as follows: Generate a random number r between [0, 1]. Based on a preset variation distribution index η m Calculate a disturbance δ according to the following formula; （10） The new parameter c after mutation is generated according to the following formula. ’ j : （11） Among them, UB j and LB j It is parameter p j The upper and lower bounds of the search, if c ’ j If it goes out of bounds, set it as a boundary value.

Images