Resonant converter dual-active burst mode control method based on time domain model

Abstract

The invention belongs to the technical field of resonant converter regulation and control, and particularly relates to a resonant converter dual-active burst mode control method based on a time domain model, which comprises the following steps: S0, dividing a burst mode into a zero state stage and a pulse packet stage; s1, establishing a time domain model of the resonant converter under dual active modulation; s2, calculating a modulation data table of the dual-active common mode in the common mode; s3, calculating regulation and control data of the corresponding starting sub-stage and stopping sub-stage; s4, generating a pulse packet lookup table of a burst mode; s5, setting a mode switching condition under closed-loop control; s6, judging whether a burst mode switching condition is met or not in a common mode, if yes, switching to the burst mode, and matching a corresponding regulation and control data packet from the pulse packet lookup table to carry out corresponding regulation and control; and in the burst mode, judging whether the condition of switching back to the common mode is met, and switching to the common mode if the condition is met. According to the method, the transient process of the burst mode under the dual-active architecture can be accurately controlled.

Description

Technical Field

[0001] The present invention belongs to the technical field of resonant converter control, and in particular relates to a dual-active burst mode control method for a resonant converter based on a time domain model. Background Art

[0002] Resonant converters are widely used in power electronics due to their soft-switching characteristics and high efficiency. However, they suffer from a significant drop in efficiency under light-load conditions. To improve light-load efficiency, burst mode control has become a key technology. Currently, the mainstream solution uses a single-active burst mode circuit—applying a pulsed drive signal only to the primary full-bridge, while the secondary side relies on the diodes to conduct naturally.

[0003] Existing implementation methods mainly include: ① Trajectory control method: By sampling the resonant cavity current / voltage signal at high frequency, the burst period is adjusted based on the time-domain trajectory control principle. However, this method requires an ultra-high-precision analog-to-digital converter (ADC) to capture the high-frequency signal in real time, which places strict demands on the controller's computing power and hardware costs; ② Dual-pulse method: Only two resonant frequency pulses are output per burst period. Although the control is simple, the energy transfer between pulses is insufficient, resulting in large current ripple, high switching losses, and limited overall efficiency; ③ Triple-pulse method: A preset pulse is added before the dual pulse to suppress current surges. Although the transient process is smoother, due to the fixed number of pulses, the problem of conduction loss and magnetic loss superposition still exists under light load.

[0004] All of the above solutions are limited by a single-active control structure: the secondary side lacks active regulation capabilities, resulting in the voltage gain being clamped to a fixed range (approximately M≈1.0). When the converter needs to operate in deep buck (e.g., M<0.5) or boost (M>1.5), the forced conduction of the secondary side diode will cause significant reverse recovery losses, and the primary side cannot independently optimize light-load efficiency. At the same time, the discontinuity of the pulse train exacerbates switching transient stress (such as current overshoot during the start / stop phase), which not only reduces efficiency but also threatens device reliability. To overcome the limitations of a single active, a dual-active (primary-secondary coordinated switching) architecture is required. This architecture can expand the voltage gain range and achieve zero current switching (ZCS) by independently controlling the two bridge arms. However, precise control of transient processes becomes a key difficulty: the initial state of the resonant cavity during the start / stop phase is strongly coupled with the quasi-steady state. If the modeling is inaccurate or the regulation is mismatched, it will lead to: 1. Transient current oscillations - causing additional switching losses and electromagnetic interference (EMI); 2. Energy transfer interruption - causing output voltage fluctuations; 3. Mode switching instability - the system oscillates between normal mode and burst mode.

[0005] Therefore, how to achieve precise control of transient processes in burst mode, wide-range voltage gain regulation, and light-load efficiency maximization under a dual-active architecture has become an urgent problem to be solved. Summary of the Invention

[0006] In response to the above-mentioned deficiencies in the prior art, the present invention provides a dual-active burst mode control method for a resonant converter based on a time domain model, which can achieve precise control of the transient process of the burst mode, wide-range adjustment of the voltage gain, and maximization of light-load efficiency under a dual-active architecture.

[0007] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0008] A dual-active burst mode control method for a resonant converter based on a time domain model comprises the following steps:

[0009] S0, the burst mode is divided into zero state stage and pulse packet stage; in the zero state, all currents and voltages in the resonant cavity are zero, and the bridge arm voltages on both sides are v ab and v cd is zero;

[0010] The pulse packet phase includes three sub-phases: the start-up sub-phase, the quasi-steady-state sub-phase, and the stop sub-phase. The quasi-steady-state is a periodic pulse. The start-up sub-phase is the intermediate process of the converter switching from zero state to quasi-steady state. The stop sub-phase is the intermediate process of the converter switching from quasi-steady state to zero state.

[0011] S1. Establish a time domain model of the resonant converter under dual active modulation;

[0012] S2. Calculating a modulation data table of a dual-active normal mode in normal mode based on a time domain model, wherein the modulation data table stores modulation data corresponding to each voltage gain m;

[0013] S3. Using each modulation data in the modulation data table as the quasi-steady-state sub-stage control data corresponding to the voltage gain m, and combining it with the time domain model, calculating the corresponding control data for the start sub-stage and the stop sub-stage;

[0014] S4. Integrate the control data of the quasi-steady-state sub-phase, the start sub-phase, and the stop sub-phase under the same voltage gain m as corresponding control data packets; generate a burst mode pulse packet lookup table based on the control data packets of each voltage gain m;

[0015] S5. Setting a mode switching condition under closed-loop control, wherein the mode switching condition includes a condition for switching to a burst mode and a condition for switching back to a normal mode;

[0016] S6. When the converter is actually running, in normal mode, determine whether the conditions for switching to burst mode are met. If so, switch to burst mode, and match the corresponding control data packet from the pulse packet lookup table based on the current voltage gain m to adjust the transformer accordingly; in burst mode, determine whether the conditions for switching back to normal mode are met. If so, switch the converter to normal mode.

[0017] Compared with the prior art, the present invention has the following beneficial effects:

[0018] 1. Achieve efficient operation over a wide voltage gain range. This method breaks through the limitation of the traditional single-active burst mode voltage gain being clamped at M≈1.0. Through dual-active collaborative modulation (active control of the primary / secondary full-bridge), it supports a wide range of voltage gain adjustment from M<0.5 (deep buck) to M>1.5 (deep boost). In the existing technology, the passive conduction of the secondary side in the single-active solution leads to reverse recovery loss and a narrow voltage gain range, which cannot be adapted to wide voltage input scenarios such as photovoltaics / energy storage. This solution uses a dual-active architecture to accurately control the bidirectional flow of energy, significantly broadening the application scenarios.

[0019] 2. Precise control of transient processes to suppress losses. Through the coordinated regulation of the start / stop sub-stage and the quasi-steady state (based on pre-calculation of the time domain model), the continuous and smooth transition of the resonant cavity current during mode switching is ensured, eliminating the current overshoot and oscillation phenomena in traditional solutions. In existing technologies, the single active burst mode causes a sudden change in energy during the start / stop stage due to the fixed number of pulses (such as the double-pulse / triple-pulse method), causing a surge in switching losses and worsening EMI. This solution optimizes the transient path and concentrates the switching losses in the zero current / zero voltage state (ZCS / ZVS), thereby improving energy efficiency.

[0020] 3. Maximize light-load efficiency. In burst mode, the quasi-steady-state sub-stage uses a periodic pulse sequence to accurately match the load demand, while the zero-state stage completely shuts off the switch tube to eliminate conduction losses, thereby minimizing losses at light loads. In the existing technology, single-active burst mode generates conduction losses due to the freewheeling of the secondary-side diode, and invalid pulses in the burst cycle lead to accumulated core losses. This solution combines dual-active modulation with segmented control to reduce invalid energy circulation and significantly improve light-load efficiency.

[0021] 4. Simplified control algorithm and enhanced practicality. This method is based on a pre-generated pulse packet lookup table (storing data packets for each gain point control). During operation, precise control can be triggered by directly matching the voltage gain parameters, without the need for real-time high-frequency sampling of the resonant cavity signal. In the existing technology, traditional trajectory control schemes rely on high-speed ADCs to sample the resonant current / voltage, which is hardware-costly and susceptible to noise interference. This method only requires sampling the input / output voltage, resulting in a simple and reliable control structure, reducing system complexity and cost.

[0022] In summary, this approach, through time-domain modeling and dual-active coordinated control, achieves precise control of burst-mode transients, dynamic adjustment of voltage gain over a wide range, and maximized light-load efficiency. Compared to single-active solutions, this significantly improves efficiency and requires only basic voltage sampling for stable operation, offering both high practicality and engineering value.

[0023] Preferably, the voltage gain m=nV2 / V1; wherein V1 is the input voltage, V2 is the output voltage, and n is the transformer ratio.

[0024] With such a setting, the working efficiency of the present method is significantly higher than that of the existing single-active method, and only the input and output voltages need to be sampled, which makes the method highly practical.

[0025] Preferably, in S2, the modulation data includes the normalized switching frequency f n , the original secondary side external phase shift D0 and the secondary side internal phase shift D2; the buck mode f n , D0 and the phase shift D1 within the primary side.

[0026] This setup achieves precise control of bidirectional energy flow. By differentiating the internal phase shift objects in the boost / buck mode (D2 acting on the secondary side, D1 acting on the primary side), it adapts to the modulation requirements of different power flows and significantly improves the gain accuracy in deep boost / buck scenarios (such as m>1.5 or m<0.5). Traditional single-active solutions have significant conduction losses at non-rated gain points (m≠1.0) because the secondary side cannot actively control the internal phase shift. This method expands the boundaries of the efficient working area through patterned phase shift distribution.

[0027] 2. Optimize switching losses and EMI performance. The internal phase shift parameters (D1 / D2) work in conjunction with the external phase shift D0 to control the soft switching timing of the switch tube, ensuring zero voltage switching (ZVS) or zero current switching (ZCS) on both the primary and secondary sides over a wide gain range, thereby suppressing diode reverse recovery losses. Existing burst mode uses a fixed pulse sequence (such as the double-pulse method). The lack of internal phase shift leads to dead time mismatch and increased switching losses at light loads. This solution dynamically adjusts the phase shift angle, significantly reducing switching losses.

[0028] 3. Reduce control complexity and storage overhead. Only three core parameters (f n ,D0,D1 / D2), simplifying the lookup table structure and improving the efficiency of real-time access to the modulation data table. Mainstream solutions require redundant parameters for the full gain range (e.g., 4-5 variables / operating conditions), which increases memory requirements and is prone to parameter conflicts. This method covers all operating conditions with a minimal set of parameters, improving system robustness.

[0029] Preferably, in S2, a preset CSO-VF-EPS modulation method is used to calculate the modulation data corresponding to each voltage gain m; the CSO-VF-EPS modulation method is to use a particle swarm optimization PSO algorithm to search for the optimal values of D0 and D1 / D2 so that the effective current value is minimized under the conditions of ZVS operation of all zero voltage turn-on switches.

[0030] This setup 1. Maximizes switching loss suppression and energy efficiency. The PSO algorithm precisely searches for the global optimal solution for D0 and D1 / D2, minimizing the effective value of the resonant current while ensuring 100% ZVS for the full-bridge switches. This reduces conduction and core losses at the source. Traditional trial-and-error methods or fixed-rule modulation (such as a fixed phase shift angle) cannot simultaneously achieve both ZVS and current minimization. Current ripple at light loads causes additional losses. This method significantly improves overall energy efficiency through intelligent optimization.

[0031] 2. Reduce real-time computing burden and hardware requirements. Optimization calculations are completed offline (S2), and the results are pre-stored in a lookup table. During actual control, parameters are directly called without running complex algorithms online, reducing the computing power required by the main control chip. Compared to existing real-time optimization solutions (such as dynamic trajectory control), which require high-frequency iterative algorithms and rely on high-end MCUs or FPGAs, this method front-loads calculations, significantly reducing system resource usage and adapting to low-cost processors.

[0032] Preferably, in buck mode, the objective function of the CSO-VF-EPS modulation method is:

[0033] Where i Lrk_rms Indicates the effective value of the dual-resonance current;

[0034] Constraints include:

[0035] Power constraint: The calculated result of power p is equal to the set value;

[0036] ZVS constraint 1: i Q_k_off t dead >2V1C oss ;

[0037] ZVS constraint 2: i S_k_off t dead >2V2C oss ;

[0038] Where, t dead is the dead time in the drive signal; i Q_k_off Indicates the primary side full bridge switch tube Q k The shutdown current; i S_k_off Indicates the secondary side full-bridge switch tube S kThe shutdown current; V1 represents the input voltage; V2 represents the output voltage; C oss Represents the output parasitic capacitance of the switching tube.

[0039] This setting can 1. Suppress the resonance loss from the source and improve the light load energy efficiency. The objective function directly optimizes the dual resonance current effective value (i Lrk_rms ), significantly reducing conduction losses and transformer core losses by minimizing the sum of squared currents, particularly improving efficiency under light-load conditions. Conventional single-active burst mode methods ignore current waveform optimization (such as the fixed three-pulse method), resulting in excessively high resonant current peaks and effective values. This solution uses mathematical modeling to precisely suppress ineffective current components, significantly improving light-load efficiency.

[0040] 2. Full-bridge switch tubes achieve 100% soft switching. The ZVS constraint strictly limits the solution space of D0, D1, fn in the form of inequality, ensuring that all switch tubes on the primary and secondary sides are in the dead time t dead The charge balance is satisfied (i Q_k_off t dead >2V1C oss and i S_k_ off t dead >2V2C oss ), completely eliminating turn-on losses. Existing solutions cannot independently control ZVS due to the passive conduction of the secondary side (for example, the secondary diode hard turns off during step-down), resulting in a significant proportion of reverse recovery losses. This approach eliminates the root cause of these losses through a dual ZVS constraint linkage design.

[0041] 3. Precise power matching and enhanced dynamic response. Power constraints enforce actual output power equal to the set value. Dynamic adjustment of the phase shift angle and frequency based on the objective function achieves high-precision power tracking during load transients. Conventional open-loop burst mode output power is strongly tied to the number of pulses, requiring waiting for new pulse packets to be generated during load changes, resulting in significant response delays. This solution optimizes parameters in real time under linear constraints, significantly improving dynamic response speed.

[0042] Preferably, in boost mode, the objective function of the CSO-VF-EPS modulation method is:

[0043]

[0044] Where i Lrk_rms Indicates the effective value of the dual-resonance current;

[0045] Constraints include:

[0046] Power constraint: The calculated result of power p is equal to the set value;

[0047] ZVS constraint 1: i Q_k_off t dead >2V1Coss ;

[0048] ZVS constraint 2: i S_k_off t dead >2V2C oss ;

[0049] Where, t dead is the dead time in the drive signal; i Q_k_off Indicates the primary side full bridge switch tube Q k The shutdown current; i S_k_off Indicates the secondary side full-bridge switch tube S k The shutdown current; V1 represents the input voltage; V2 represents the output voltage; C oss Represents the output parasitic capacitance of the switching tube.

[0050] Preferably, in S3, the promoter stage is activated by regulating v ab and v cd The waveform of the startup phase is used to indirectly control the state variables, so that the values of all state variables at the end of the startup phase are approximately equal to the values at the initial moment of the quasi-steady-state phase; and the duration of the startup phase is t c Minimization is the optimization goal, and the PSO algorithm is used to solve it;

[0051] The objective function is: min t c ;

[0052] Constraints include:

[0053] Continuity constraint: x(t c )=x′(T s / 2-t d )

[0054] ZVS constraint: The number of non-ZVS switches does not exceed 1;

[0055] Among them, x′(T s / 2-t d ) is the expected initial value of the state variable in the quasi-steady state; t d represents the duration of the first switching cycle of the switch to the quasi-steady-state sub-phase; T s represents the duration of a complete switching cycle in the quasi-steady-state sub-phase; x(t c ) represents the state variables at the end of the initial stage.

[0056] This setting can achieve high-speed and smooth switching of transient processes. c Minimization is the goal, and PSO is used to dynamically optimize v ab / v cdThe waveform compresses the startup time to the theoretical limit, significantly speeding up the startup strategy compared to the traditional fixed-time startup strategy and eliminating the risk of current overshoot. Conventional burst mode uses a fixed number of pulses to start (such as presetting 3 cycle pulses), which takes a long time and is decoupled from the load; this solution is based on the state variable trajectory prediction (x(t c )→x′), realizing fast switching of load adaptation.

[0057] 2. Ensure seamless connection of quasi-steady state and suppress transient loss. Continuity constraint forces state variables to be constant at t c Matching the endpoint with the quasi-steady-state initial value avoids current / voltage steps at the switching moment, significantly reducing transient oscillation losses. In existing single-active solutions, due to the uncontrollable secondary side, the state variables at the end of startup randomly deviate from the target value (such as current offset caused by diode conduction), causing subsequent pulse sequence instability. This strategy improves the energy transfer efficiency of the pulse packet through precise state alignment.

[0058] 3. Balancing soft-switching requirements with algorithm robustness. ZVS fault-tolerant constraints (non-ZVS ≤ 1) allow some switches to sacrifice soft-switching characteristics in exchange for PSO solution feasibility and computational efficiency. Compared to strict ZVS constraints, this improves optimization speed and convergence success rate. Traditional methods that insist on full ZVS may result in no solution (e.g., insufficient secondary current during deep buck). This solution adapts to boundary conditions through fault-tolerant design, expanding the scope of the control strategy.

[0059] Preferably, in S3, compared with the start sub-stage, when calculating the state variables in the stop sub-stage, the time axis is reversed, and the rest of the calculation process is the same.

[0060] This means that a , t b and t c will take a negative value;

[0061] Preferably, in S6, in normal mode, the PI regulator 1 is used to dynamically adjust the switching frequency f s , so that the output voltage V2 is at the reference value V 2_ref Control the target; the switching burst mode condition is f s ≥f s_max , where f s_max Indicates the preset switching frequency threshold.

[0062] This setting can achieve load adaptation and steady-state accuracy improvement. The PI regulator 1 can dynamically adjust f in real time. s , automatically matches the load changes to ensure that the output voltage V2 accurately follows the reference value V 2_refThis avoids the static errors of traditional fixed-frequency control, even when the input voltage fluctuates or the load changes. Traditional open-loop burst mode relies solely on a preset pulse sequence and cannot respond to dynamic load changes, resulting in output voltage overshoot / undershoot at light loads. This solution incorporates closed-loop regulation to improve steady-state accuracy.

[0063] 2. Optimize the mode switching boundary to suppress invalid loss. s ≥f s_max As the only condition for switching to burst mode, switching is triggered only when light load is needed to improve efficiency, avoiding frequent mode oscillations caused by misjudgment of voltage threshold in traditional solutions. Existing methods often use output voltage threshold to trigger burst mode (such as V2>V 2_th ), which is susceptible to noise interference, causing false switching and increasing switching losses; this scheme uses the frequency threshold as the criterion, which has clear physical meaning and strong anti-interference ability.

[0064] 3. Reduced closed-loop control complexity. Only PI regulator 1 is required to control fs (no resonant current sampling is required), simplifying the control structure and reducing computational resource consumption. Existing high-performance trajectory control schemes require multivariable feedback (resonant current / voltage), relying on high-speed ADCs and complex algorithms. This method uses only a basic voltage signal and a frequency regulator, resulting in cost-effectiveness.

[0065] Preferably, in S6, in burst mode, the burst frequency f is dynamically adjusted using the PI regulator 2 bur , so that the output voltage V2 is at the reference value V 2_ref Control the target; the condition for switching back to normal mode is f bur ≥f bur_max , where f bur_max Indicates the preset burst frequency threshold. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to make the purpose, technical solutions and advantages of the invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings, in which:

[0067] Figure 1 Flowchart of this method;

[0068] Figure 2 1 is a main circuit diagram of a CLLC resonant converter in an embodiment;

[0069] Figure 3 Schematic diagram of the operating waveform of the CLLC converter under VF-EPS modulation in the buck mode in the embodiment;

[0070] Figure 4 Schematic diagram of the operating waveform of the CLLC converter under VF-EPS modulation in boost mode in the embodiment;

[0071] Figure 5Schematic diagram of the equivalent circuit of the CLLC resonant cavity in the embodiment;

[0072] Figure 6 Schematic diagram of the change of the sum of squares of effective current values with modulation variables D0 and D1 in the embodiment;

[0073] Figure 7 Schematic diagram of control variable optimization results when p=500W in the embodiment;

[0074] Figure 8 Schematic diagram of dual-active burst mode operating waveform in an embodiment;

[0075] Figure 9 Schematic diagram of the waveform of the startup state when the voltage gain m=0.75 in the embodiment;

[0076] Figure 10 : is a closed-loop control block diagram in an embodiment;

[0077] Figure 11 Schematic diagram of the full power range control rate in the embodiment;

[0078] Figure 12 This is a diagram of the actual working waveform in the embodiment. DETAILED DESCRIPTION

[0079] The following is a further detailed description through specific implementation methods:

[0080] Example:

[0081] like Figure 1 As shown, this embodiment discloses a dual-active burst mode control method for a resonant converter based on a time domain model, comprising the following steps:

[0082] S0, the burst mode is divided into zero state stage and pulse packet stage; in the zero state, all currents and voltages in the resonant cavity are zero, and the bridge arm voltages on both sides are v ab and v cd is zero;

[0083] The pulse packet phase includes three sub-phases: the start-up sub-phase, the quasi-steady-state sub-phase, and the stop sub-phase. The quasi-steady-state is a periodic pulse. The start-up sub-phase is the intermediate process of the converter switching from zero state to quasi-steady state. The stop sub-phase is the intermediate process of the converter switching from quasi-steady state to zero state.

[0084] S1. Establish a time domain model of a resonant converter under dual active modulation; the circuit structure of the resonant converter includes a primary side full bridge, a secondary side full bridge and a middle resonant cavity.

[0085] In order to facilitate those skilled in the art to better understand the technical content of S1, a CLLC resonant converter is taken as an example for illustration.

[0086] The resonant converter has good zero voltage switching (ZVS) soft switching performance. The main circuit diagram of the CLLC resonant converter is as follows: Figure 2 As shown in the figure, the circuit structure can be divided into three parts: the primary side full bridge, the middle resonant cavity and the secondary side full bridge. The difference between other types of resonant converters such as SRC, LLC resonant converter and CLLC resonant converter lies in the different structures of the resonant cavity. The resonant cavity of the CLLC resonant converter has a higher circuit order, so CLLC is used as an example here. Other resonant converters can be applied to this method by directly adjusting the input CLLC circuit parameters, such as letting L r2 Equal to 0. C r is equal to infinity, then CLLC becomes LLC, and further L m If set to infinity, it becomes SRC again.

[0087] The working waveform of the CLLC resonant converter under VF-EPS modulation is as follows: Figure 3 and Figure 4 As shown, compared with Extended Phase Shift (EPS), the D0, D2 and f n In addition, the normalized switching frequency f n =f s / f r As a new control variable. Together they determine v ab and v cd The waveform of the converter switches between different levels at different moments. According to these switching moments, the converter's operating waveform can be divided into different modes.

[0088] The equivalent circuit of a single mode is as follows Figure 5 As shown, where v ab and v cd is considered a constant. In addition, there are four independent state variables:

[0089] x=[i Lr1 i Lr2 / n -v Cr1 nv Cr2 ] T (1)

[0090] According to Kirchhoff's current law (KCL) and Kirchhoff's voltage law (KVL), we have

[0091]

[0092] According to the differential equations of capacitors and inductors, we can get

[0093]

[0094] Combining Equation (2) and Equation (3) and eliminating the state variables other than vCr2, we can obtain

[0095]

[0096] in

[0097]

[0098] The characteristic equation of formula (4) has two pairs of conjugate imaginary roots, represented by ±ω1 and ±ω2, where

[0099]

[0100] Then, we can get n(v Cr2 +v cd ) and its derivatives are

[0101] D(t)=M C (t)C (7)

[0102] in

[0103]

[0104] D(t)=[n(v Cr2 +v cd ) nv′ Cr2 nv″ Cr2 nv″′ Cr2 ] T (9)

[0105] and

[0106] C=[c1 c2 c3 c4] T (10)

[0107] Combining equations (2) and (3), we can also get

[0108] D(t)=M V (x(t)+S) (11)

[0109] in

[0110]

[0111] S=[0 0 v ab nv cd ] T (13)

[0112] Combining (7) and (11) we can get:

[0113] C=M C (0) -1 D(0)=M C (0) -1 M V (x(0)+S) (14)

[0114] and

[0115]

[0116] The quantities on the right side of the equal sign in Equation (15) are all known except x(0). Therefore, for any mode, only x(0) is needed to obtain x(t) at any time through Equation (15).

[0117] According to the symmetry of the state variables in the half switching cycle in the steady state, Figure 2 and Figure 3 The initial values of the state variables can be expressed as

[0118]

[0119] Among them S 1-3 Represent the bridge arm voltage vectors at each stage respectively. By transforming Equation (16), we can get the expression of x(0) as

[0120]

[0121] At this point, the state variable value at any time within the switching cycle can be calculated.

[0122] S2. Based on the time domain model, a modulation data table of the dual-active normal mode in the normal mode is calculated, where the modulation data table stores modulation data corresponding to each voltage gain m.

[0123] Wherein, voltage gain m=nV2 / V1;wherein, V1 is input voltage, V2 is output voltage, and n is transformer ratio. The modulation data includes normalized switching frequency f of boost mode. n , the original secondary side external phase shift D0 and the secondary side internal phase shift D2; the buck mode f n , D0 and the phase shift D1 in the primary side. n =f s / f r ; Among them, f s is the original switching frequency, f r is the resonant frequency.

[0124] During specific implementation, a preset CSO-VF-EPS modulation method is used to calculate modulation data corresponding to each voltage gain m.

[0125] Taking the above example, VF-EPS has three degrees of freedom of modulation (D0, D2 and f in boost mode). n and D0, D1 and f in buck mode n ), taking the buck mode as an example, under the condition of a given voltage gain m, adjust f n Let the power be 500w. At this time, the effective value of the current changes with D0 and D1 as follows: Figure 6 shown.

[0126] The CSO-VF-EPS modulation method uses a particle swarm optimization (PSO) algorithm to search for optimal values for D0 and D1 / D2 to minimize the RMS current under ZVS conditions for all switches. This PSO algorithm accurately searches for the global optimal solution for D0 and D1 / D2, minimizing the RMS resonant current while ensuring 100% ZVS for all full-bridge switches, thereby reducing conduction and core losses. Traditional trial-and-error methods or fixed-rule modulation (such as a fixed phase shift angle) cannot simultaneously achieve ZVS and current minimization, resulting in additional losses due to current ripple at light loads. This method significantly improves overall energy efficiency through intelligent optimization. Furthermore, the optimization calculations are performed offline (S2), and the results are pre-stored in a lookup table. During actual control, the parameters are directly called up, eliminating the need for complex online algorithms and reducing the computing power requirements of the main control chip. Compared to existing real-time optimization solutions (such as dynamic trajectory control), which require high-frequency iterative algorithms and rely on high-end MCUs or FPGAs, this method front-ends the calculations, significantly reducing system resource usage and making it compatible with low-cost processors.

[0127] In buck mode, the objective function of the CSO-VF-EPS modulation method is:

[0128] Where i Lrk rms Indicates the effective value of the dual-resonance current;

[0129] Constraints include:

[0130] Power constraint: The calculated result of power p is equal to the set value;

[0131] ZVS constraint 1: i Q_k_off t dead >2V1C oss ;

[0132] ZVS constraint 2: i S_k_off t dead >2V2C oss ;

[0133] Where, t dead is the dead time in the drive signal; i Q_k_off Indicates the primary side full bridge switch tube Qk The shutdown current; i S_k_off Indicates the secondary side full-bridge switch tube S k The shutdown current; V1 represents the input voltage; V2 represents the output voltage; C oss Represents the output parasitic capacitance of the switching tube.

[0134] The p in the power constraint is set to 50% rated load, because this is the highest efficiency point of CSO-VF-EPS modulation. Then, the optimal results for different m are as follows Figure 7 shown.

[0135] In this way, the objective function directly optimizes the effective value of the dual-resonance current (i Lrk_rms ), by minimizing the sum of squared currents, significantly reducing conduction losses and transformer core losses, especially improving light-load efficiency. Traditional single-active burst mode ignores current waveform optimization (such as the fixed three-pulse method), resulting in excessively high resonant current peak / effective value. This solution accurately suppresses the ineffective current component through mathematical modeling, significantly improving light-load efficiency. In addition, the ZVS constraint strictly limits the solution space of D0, D1, and fn in the form of an inequality, ensuring that all switches on the primary and secondary sides are within the dead time t dead The charge balance is satisfied (i Q_k_off t dead >2V1C oss and i S_k_off t dead >2V2C oss ), completely eliminating turn-on losses. The existing solution cannot independently control ZVS due to the passive conduction of the secondary side (such as the hard turn-off of the secondary side diode during voltage reduction), and the reverse recovery loss accounts for a large proportion; this method eliminates the root cause of the loss through the dual ZVS constraint linkage design. In addition, it can achieve precise power matching and enhanced dynamic response. The power constraint forces the actual output power to be equal to the set value, and dynamically adjusts the phase shift angle and frequency in combination with the objective function to achieve high-precision power tracking during load transients. The output power of conventional open-loop burst mode is strongly bound to the number of pulses. When the load jumps, it is necessary to wait for the generation of a new pulse packet, and the response delay is relatively large; this solution optimizes parameters in real time under linear constraints, and the dynamic response speed is significantly improved.

[0136] In the pressure mode, the objective function of the CSO-VF-EPS modulation method is:

[0137]

[0138] Where i Lrk_rms Indicates the effective value of the dual-resonance current;

[0139] Constraints include:

[0140] Power constraint: The calculated result of power p is equal to the set value;

[0141] ZVS constraint 1: i Q_k_off t dead >2V1C oss ;

[0142] ZVS constraint 2: i S_k_off t dead >2V2C oss ;

[0143] Where, t dead is the dead time in the drive signal; i Q_k_off Indicates the primary side full bridge switch tube Q k The shutdown current; i S_k_off Indicates the secondary side full-bridge switch tube S k The shutdown current; V1 represents the input voltage; V2 represents the output voltage; C oss Represents the output parasitic capacitance of the switching tube.

[0144] Still taking the above example as an example, the dual active burst mode control strategy of the resonant converter proposed by this method is as follows Figure 8 As shown, it can be seen that there are two stages, namely the zero state stage and the pulse packet stage. In the zero state, all currents and voltages in the resonant cavity are zero, and Q 1,3 and S 1,3 Keep conducting to ensure the bridge arm voltage v ab =v cd = 0. Each pulse packet can be divided into three sub-phases: a start-up sub-phase, a quasi-stable sub-phase, and a stop sub-phase. The quasi-stable sub-phase can be any type of dual-active common mode modulation. The drive signals for the start-up and stop sub-phases are optimized using time-domain models and optimization algorithms. This fully utilizes the control freedom of dual-active modulation and enables smooth transitions from zero state to quasi-stable state and back again.

[0145] S3. Using each modulation data in the modulation data table as the quasi-steady-state sub-stage control data corresponding to the voltage gain m, and combining it with the time domain model, calculating the corresponding control data of the start sub-stage and the stop sub-stage.

[0146] During the quasi-steady-state phase, the switch's drive signal repeats several switching cycles, and power transfer primarily occurs during this phase. The number of switching cycles can be adjusted as needed. Generally speaking, fewer cycles result in lower output voltage ripple and a higher burst frequency, but also lower efficiency. Because the output voltage varies very little within a single pulse packet, the converter's performance during this phase can be considered identical to that in the steady-state phase. Therefore, the modulation scheme and transmission power can be freely selected. This method employs the aforementioned CSO-VF-EPS modulation during the quasi-steady-state phase.

[0147] The startup phase can be regarded as the intermediate process of the converter switching from zero state to quasi-steady state. The task of this phase is to ensure smooth switching. ab and v cd The waveform of the startup phase is used to indirectly control the state variables, so that the values of all state variables at the end of the startup phase are approximately equal to the values at the initial moment of the quasi-steady-state phase; and the duration of the startup phase is t c Minimization is the optimization goal, and the PSO algorithm is used to solve it;

[0148] The objective function is: min t c ;

[0149] Constraints include:

[0150] Continuity constraint: x(t c )=x′(T s / 2-t d )

[0151] ZVS constraint: The number of non-ZVS switches does not exceed 1;

[0152] Where x′(t) represents the state variable in the quasi-steady-state CSO-VF-EPS waveform, so x′(T s / 2-t d ) is the expected initial value of the state variable under quasi-steady state; t d represents the duration of the first switching cycle of the switch to the quasi-steady-state sub-phase; T s represents the duration of a complete switching cycle in the quasi-steady-state sub-phase; x(t c ) represents the state variables at the end of the initial stage.

[0153] In this way, high-speed and smooth switching of transient processes can be achieved. c Minimization is the goal, and PSO is used to dynamically optimize v ab / v cd The waveform compresses the startup time to the theoretical limit, significantly speeding up the startup strategy compared to the traditional fixed-time startup strategy and eliminating the risk of current overshoot. Conventional burst mode uses a fixed number of pulses to start (such as presetting 3 cycle pulses), which takes a long time and is decoupled from the load; this solution is based on the state variable trajectory prediction (x(t c )→x′), realizing fast switching of load adaptation. In addition, the continuity constraint forces the state variables to be cThe end point matches the quasi-steady-state initial value to avoid the current / voltage step at the switching moment, which can greatly reduce the transient oscillation loss. In the existing technology, the single-active solution is uncontrollable on the secondary side, and the state variables at the end of the startup randomly deviate from the target value (such as the current offset caused by the diode conduction), causing the subsequent pulse sequence to become unstable; this strategy improves the energy transfer efficiency of the pulse packet through precise state alignment. In addition, the ZVS fault-tolerant constraint (non-ZVS ≤ 1) allows local switches to sacrifice soft switching characteristics in exchange for the feasibility and computational efficiency of the PSO solution space. Compared with strict ZVS constraints, the optimization speed is improved and the convergence success rate is higher. Traditional methods that force full ZVS may lead to no solution (such as insufficient secondary current during deep voltage reduction); this solution adapts to boundary conditions through fault-tolerant design and expands the scope of application of the control strategy.

[0154] Still taking the above example, the waveform of the initiator stage is as follows: Figure 9 As shown, when t=t c When t=t, the resonant converter switches from the startup sub-phase to the quasi-steady-state sub-phase. c +t d is the end time of the first half switching cycle in the quasi-steady-state stage, t d The initial phase of the quasi-steady-state sub-stage is determined.

[0155] As shown in Table 1, there are eight independent variables to control v ab and v cd The waveform, where v p3 and v s3 are excluded because they are part of the waveform of the quasi-steady-state subphase, represented by t d Decide.

[0156] Table 1. Value ranges of startup phase variables

[0157] variable type Value range <![CDATA[t a ]]> continuous <![CDATA[(0,t c )]]> <![CDATA[t b ]]> continuous <![CDATA[(0,t c )]]> <![CDATA[t c ]]> continuous <![CDATA[(0,T s / 2)]]> <![CDATA[t d ]]> continuous <![CDATA[(0,T s / 2)]]> <![CDATA[v p1 ]]> Discrete <![CDATA[-V1,0,V1]]> <![CDATA[v p2 ]]> Discrete <![CDATA[-V1,0,V1]]> <![CDATA[v s1 ]]> Discrete <![CDATA[-V2,0,V2]]> <![CDATA[v s2 ]]> Discrete <![CDATA[-V2,0,V2]]>

[0158] According to equations (14) and (15), the expression of the state variable in the startup phase is:

[0159]

[0160] Non-ZVS switching significantly increases switching losses. However, the ZVS implementation during startup can be calculated based on the voltage levels and state variables before and after each mode switch. Because model errors accumulate over time, the duration of the startup phase should be minimized. This serves as the optimization objective, and the PSO algorithm is again employed for the solution.

[0161] Compared with the start sub-phase, the stop sub-phase reverses the time axis when calculating state variables, and the rest of the calculation process is the same.

[0162] The standard function is: min tc ;

[0163] Constraints include:

[0164] Continuity constraint: x(-t c )=x′(t d )

[0165] ZVS constraint: The number of non-ZVS switches does not exceed 1;

[0166] Among them, x′(t d ) is the expected terminal value of the state variable in the quasi-steady state; t d represents the duration of the last switching cycle of the quasi-steady-state sub-phase; T s represents the complete duration of the switching cycle of the quasi-steady-state sub-phase; x(-t c ) represents the initial value of the state variable at the stop sub-phase, and the state variable value x(0)=0 at the end of the stop sub-phase.

[0167] S4. Integrate the control data of the quasi-steady-state sub-phase, the start sub-phase, and the stop sub-phase under the same voltage gain m as corresponding control data packets; generate a burst mode pulse packet lookup table based on the control data packets of each voltage gain m;

[0168] S5. Setting a mode switching condition under closed-loop control, wherein the mode switching condition includes a condition for switching to the burst mode and a condition for switching back to the normal mode.

[0169] S6. When the converter is actually running, in normal mode, determine whether the conditions for switching to burst mode are met. If so, switch to burst mode, and match the corresponding control data packet from the pulse packet lookup table based on the current voltage gain m to adjust the transformer accordingly; in burst mode, determine whether the conditions for switching back to normal mode are met. If so, switch the converter to normal mode.

[0170] In the specific implementation, in normal mode, the PI regulator 1 is used to dynamically adjust the switching frequency f s , so that the output voltage V2 is at the reference value V 2_ref Control the target; the switching burst mode condition is f s ≥f s_max , where f s_max Indicates the preset switching frequency threshold. In burst mode, the burst frequency f is dynamically adjusted using PI regulator 2. bur , so that the output voltage V2 is at the reference value V 2_ref Control the target; the condition for switching back to normal mode is f bur ≥f bur_max , where f bur_max Indicates the preset burst frequency threshold.

[0171] In specific implementation, the closed-loop control block diagram of the proposed burst mode CLLC converter is as follows: Figure 10 As shown. Among them, "PI regulator 2" is used to dynamically adjust the burst frequency f bur To stabilize the output voltage V2 at the reference value V 2_ref Since the optimization result varies with the voltage gain m (e.g. Figure 7 It is necessary to establish a lookup table containing pulse packets with different voltage gains (ie, control data packets).

[0172] When each pulse packet is about to start, the data latch inside the modulator reads and stores the pulse packet data (control data packet) corresponding to the current voltage gain m. Subsequently, the pulse packet is sequentially converted into a set of drive signal sequences and input into the CLLC converter.

[0173] The control law in the entire load range is as follows Figure 11 As shown. In burst mode, f s Represents the switching frequency of the quasi-steady-state sub-stage. This frequency has nothing to do with power (because the transmission power in the quasi-steady state is fixed at 500W), and f bur Will increase with the increase of power. bur ≥f bur_max , the converter will switch to normal mode.

[0174] In normal mode, the switching frequency f s Will decrease with the increase of power p. When p decreases, f s will gradually increase until it reaches f s_max , at which point the converter switches back to burst mode. Since there is a difference in fs between the two modes, a hysteresis loop is added to the system to enhance robustness.

[0175] Still taking the above example as an example, the measured working waveform of the burst mode CLLC converter proposed by this method is as follows: Figure 12 shown.

[0176] This method breaks through the limitation of the traditional single-active burst mode voltage gain being clamped at M≈1.0. Through dual-active collaborative modulation (active control of the primary / secondary full-bridge), it supports a wide range of voltage gain adjustment from M<0.5 (deep buck) to M>1.5 (deep boost). In the existing technology, the passive conduction of the secondary side in the single-active solution leads to reverse recovery loss and a narrow voltage gain range, which cannot be adapted to wide voltage input scenarios such as photovoltaics / energy storage. This solution accurately controls the bidirectional flow of energy through a dual-active architecture, significantly broadening the application scenarios. In addition, this method ensures a continuous and smooth transition of the resonant cavity current during mode switching through the coordinated control of the start / stop sub-stage and quasi-steady state (based on pre-calculation of the time domain model), eliminating the current overshoot and oscillation phenomena in traditional solutions. In the prior art, the single-active burst mode causes a sudden change in energy during the start / stop phase due to the fixed number of pulses (such as the double-pulse / triple-pulse method), which leads to a surge in switching losses and worsening EMI. This solution optimizes the transient path to concentrate the switching losses in the zero-current / zero-voltage state (ZCS / ZVS), thereby improving energy efficiency. In the burst mode, the quasi-steady-state sub-stage uses a periodic pulse sequence to accurately match the load demand, while the zero-state stage completely turns off the switch tube to eliminate the conduction loss, thereby minimizing the loss under light load. In the prior art, the single-active burst mode generates conduction loss due to the freewheeling of the secondary-side diode, and the invalid pulses in the burst period cause the accumulation of magnetic core loss. This solution combines dual-active modulation with segmented control to reduce invalid energy circulation and significantly improve light-load efficiency. In addition, this method is based on a pre-generated pulse packet lookup table (storing the control data packets of each gain point). During operation, precise control can be triggered by directly matching the voltage gain parameters without the need for real-time high-frequency sampling of the resonant cavity signal. In the existing technology, traditional trajectory control schemes rely on high-speed ADCs to sample the resonant current / voltage, which has high hardware costs and is susceptible to noise interference. This method only requires sampling the input / output voltage, has a simple and reliable control structure, and reduces system complexity and cost.

[0177] This method adapts to the modulation requirements of different power flows by distinguishing the internal phase shift objects in the boost / buck mode (D2 acts on the secondary side, D1 acts on the primary side), significantly improving the gain accuracy in deep boost / buck scenarios (such as m>1.5 or m<0.5). The traditional single-active solution has significant conduction losses at non-rated gain points (m≠1.0) because the secondary side cannot actively control the internal phase shift; this method expands the boundaries of the efficient working area through patterned phase shift distribution. In addition, the internal phase shift parameters (D1 / D2) cooperate with the external phase shift D0 to control the soft switching timing of the switch tube, ensuring that the primary / secondary side achieves zero voltage switching (ZVS) or zero current switching (ZCS) under a wide gain range, thereby suppressing the reverse recovery loss of the diode. The existing burst mode uses a fixed pulse sequence (such as the double-pulse method). The lack of internal phase shift leads to dead time mismatch, which increases switching losses at light loads; this solution dynamically adjusts the phase shift angle, significantly reducing switching losses. In addition, only three core parameters need to be stored for each operating condition (fn ,D0,D1 / D2), simplifying the lookup table structure and improving the efficiency of real-time access to the modulation data table. Mainstream solutions require redundant parameters for the full gain range (e.g., 4-5 variables / operating conditions), which increases memory requirements and is prone to parameter conflicts. This method covers all operating conditions with a minimal set of parameters, improving system robustness.

[0178] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the technical solutions. Those skilled in the art should understand that modifications or equivalent replacements of the technical solutions of the present invention that do not depart from the purpose and scope of the technical solutions of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A dual-active burst mode control method for a resonant converter based on a time domain model, characterized in that: The following steps are involved: S0, the burst mode is divided into zero state stage and pulse packet stage; In the zero state, all currents and voltages in the resonant cavity are zero, and the bridge arm voltages on both sides are v ab and v cd is zero; the pulse packet phase includes three sub-phases, the start sub-phase, the quasi-steady-state sub-phase and the stop sub-phase; the quasi-steady-state is a periodic pulse, the start sub-phase is the intermediate process of the converter switching from the zero state to the quasi-steady state, and the stop sub-phase is the intermediate process of the converter switching from the quasi-steady state to the zero state; S1. Establish a time domain model of the resonant converter under dual active modulation; S2. Calculating a modulation data table of a dual-active normal mode in normal mode based on a time domain model, wherein the modulation data table stores modulation data corresponding to each voltage gain m; S3. Using each modulation data in the modulation data table as the quasi-steady-state sub-stage control data corresponding to the voltage gain m, and combining it with the time domain model, calculating the corresponding control data for the start sub-stage and the stop sub-stage; S4, integrating the control data of the quasi-steady-state sub-phase, the start sub-phase, and the stop sub-phase under the same voltage gain m as a corresponding control data packet; Generate a burst mode pulse packet lookup table based on the control data packets of each voltage gain m; S5. Setting a mode switching condition under closed-loop control, wherein the mode switching condition includes a condition for switching to a burst mode and a condition for switching back to a normal mode; S6. When the converter is actually running, in normal mode, determine whether the conditions for switching to burst mode are met. If so, switch to burst mode, and match the corresponding control data packet from the pulse packet lookup table based on the current voltage gain m to adjust the transformer accordingly; in burst mode, determine whether the conditions for switching back to normal mode are met. If so, switch the converter to normal mode.

2. The dual-active burst mode control method for a resonant converter based on a time domain model according to claim 1, wherein: Voltage gain m = nV2 / V1; where V1 is the input voltage, V2 is the output voltage, and n is the transformer ratio.

3. The dual-active burst mode control method for a resonant converter based on a time domain model according to claim 2, wherein: In S2, the modulation data includes the normalized switching frequency f in the boost mode. n , the original secondary side external phase shift D0 and the secondary side internal phase shift D2; the buck mode f n , D0 and the phase shift D1 within the primary side.

4. The dual-active burst mode control method for a resonant converter based on a time domain model according to claim 3, wherein: In S2, a preset CSO-VF-EPS modulation method is used to calculate the modulation data corresponding to each voltage gain m. The CSO-VF-EPS modulation method uses a particle swarm optimization (PSO) algorithm to search for the optimal values of D0 and D1 / D2 so as to minimize the effective current value under the condition that all zero voltage switching tubes (ZVS) operate.

5. The dual-active burst mode control method for a resonant converter based on a time domain model according to claim 4, wherein: In buck mode, the objective function of the CSO-VF-EPS modulation method is: Where i Lrk_rms Indicates the effective value of the dual-resonance current; Constraints include: Power constraint: The calculated result of power p is equal to the set value; ZVS constraint 1: i Q_k_off t dead >2V1C oss ; ZVS constraint 2: i S_k_off t dead >2V2C oss ; Where, t dead is the dead time in the drive signal; i Q_k_off Indicates the primary side full bridge switch tube Q k The shutdown current; i S_k_off Indicates the secondary side full-bridge switch tube S k The shutdown current; V1 represents the input voltage; V2 represents the output voltage; C oss Represents the output parasitic capacitance of the switching tube.

6. The dual-active burst mode control method for a resonant converter based on a time domain model according to claim 5, wherein: In boost mode, the objective function of the CSO-VF-EPS modulation method is: Where i Lrk_rms Indicates the effective value of the dual-resonance current; Constraints include: Power constraint: The calculated result of power p is equal to the set value; ZVS constraint 1: i Q_k_off t dead >2V1C oss ; ZVS constraint 2: i S_k_off t dead >2V2C oss ; Where, t dead is the dead time in the drive signal; i Q_k_off Indicates the primary side full bridge switch tube Q k The shutdown current; i S_k_off Indicates the secondary side full-bridge switch tube S k The shutdown current; V1 represents the input voltage; V2 represents the output voltage; C oss Represents the output parasitic capacitance of the switching tube.

7. The dual-active burst mode control method for a resonant converter based on a time domain model according to claim 6, wherein: In S3, the promoter stage regulates v ab and v cd The waveform of the startup phase is used to indirectly control the state variables, so that the values of all state variables at the end of the startup phase are approximately equal to the values at the initial moment of the quasi-steady-state phase; and the duration of the startup phase is t c Minimization is the optimization goal, and the PSO algorithm is used to solve it; The objective function is: min t c ; Constraints include: Continuity constraint: x(t c )=x′(T s / 2-t d ) ZVS constraint: The number of non-ZVS switches does not exceed 1; Among them, x′(T s / 2-t d ) is the expected initial value of the state variable in the quasi-steady state; t d represents the duration of the first switching cycle of the switch to the quasi-steady-state sub-phase; T s represents the duration of a complete switching cycle in the quasi-steady-state sub-phase; x(t c ) represents the state variables at the end of the initial stage.

8. The dual-active burst mode control method for a resonant converter based on a time domain model according to claim 7, wherein: In S3, the stop sub-phase is different from the start sub-phase in that the time axis is reversed when calculating the state variables, and the rest of the calculation process is the same.

9. The dual-active burst mode control method for a resonant converter based on a time domain model according to claim 8, wherein: In S6, in normal mode, PI regulator 1 is used to dynamically adjust the switching frequency f s , so that the output voltage V2 is at the reference value V 2_ref Control the target; the switching burst mode condition is f s ≥f s_max , where f s_max Indicates the preset switching frequency threshold.

10. The dual-active burst mode control method for a resonant converter based on a time domain model according to claim 9, wherein: In S6, in burst mode, PI regulator 2 is used to dynamically adjust the burst frequency f bur , so that the output voltage V2 is at the reference value V 2_ref Control the target; the condition for switching back to normal mode is f bur ≥f bur_max , where f bur_max Indicates the preset burst frequency threshold.

Images