A Loss Model-Based Efficiency Optimization Control Method and System for DAB Micro-Inverters

Abstract

This invention discloses an efficiency optimization control method and system for a dual active bridge micro-inverter based on a loss model. The method first establishes a system model for frequency conversion linearization control, defining the phase shift coefficient as the ratio of the inner and outer phase shift angles and ensuring a linear relationship between the output current and the inner phase shift angle. Then, it establishes an accurate loss model including conduction losses, switching losses, transformer winding losses, and core losses. With the goal of minimizing the average total loss over the power frequency cycle, and using the phase shift coefficient as the optimization variable, an optimization algorithm is used to perform offline global optimization across all operating conditions, generating an optimal phase shift coefficient lookup table covering all operating conditions. During online operation, photovoltaic voltage and output power are collected in real time, the optimal phase shift coefficient is obtained by querying the lookup table, and extended phase shift modulation is used to generate a pulse width modulation signal. This invention achieves adaptive optimization of total losses across all operating conditions by combining offline global optimization with online table lookup. It has a low computational burden, is easy to implement in low-cost controllers, and significantly improves inverter operating efficiency.

Description

Technical Field

[0001] This invention relates to the field of power electronic converter control technology, and in particular to a method and system for optimizing the efficiency control of DAB micro-inverters based on a loss model. Background Technology

[0002] Half-bridge dual active bridge (DAB) microinverters have gained widespread attention in the field of distributed photovoltaic power generation due to their advantages such as electrical isolation, bidirectional power flow, and ease of soft switching. This topology transmits energy through a high-frequency transformer and leakage inductance, and its control strategy often employs extended phase-shift modulation (EPS), introducing an internal phase shift angle on the primary side of the full bridge. The system has two degrees of freedom: the external phase shift angle φ. To simplify control and achieve linear regulation of the output current, existing technologies have proposed frequency conversion-based linearization control strategies, which define a phase shift coefficient. And the specific functional relationship between the switching frequency and the inner phase shift angle, so that the output current and the control variable... The linear relationship effectively reduces the difficulty of controller design.

[0003] However, the value of the phase shift coefficient k has a crucial impact on system performance. It not only determines the magnitude of reactive power return current and current stress, but also directly alters the soft-switching range and conduction loss distribution of the power switching transistors. The system's loss composition changes significantly over a wide input voltage and load range. Under light loads, switching losses account for a high proportion, making full-range zero-voltage switching (ZVS) crucial for efficiency. Under heavy loads, conduction losses become dominant, making it more important to reduce the effective value of the current and reactive circulating current.

[0004] In the prior art, some representative solutions have emerged for efficiency optimization of DAB microinverters. For example, Chinese invention patent CN118646276A (publication date 2026.03.17) discloses an efficiency optimization method based on frequency conversion phase-shift modulation of DAB converters. This method calculates the per-unit value of output power, divides the operating modes, solves the limiting power, and plans the optimal phase-shift combination (first phase-shift, second phase-shift, and third phase-shift) and switching frequency under different operating conditions to reduce current stress and RMS current while achieving zero-voltage conduction across the entire power range. However, this scheme still has the following shortcomings: First, its optimization objective is based on the premise of "achieving zero-voltage conduction of all switches across the full power range," and on this basis, it reduces current stress and effective current value. The focus of optimization is on reducing a single electrical stress index, rather than directly minimizing the total converter loss. Second, this scheme uses multiple shift ratios as control variables and determines the parameter values ​​for each operating condition through mode division and analytical solution of the limit power. Essentially, it is a piecewise optimization scheme based on analytical derivation, which cannot achieve smooth and continuous optimization across the entire operating range. Third, the loss analysis of this scheme mainly focuses on the switching and conduction losses of the switches, without systematically considering the high-frequency eddy current losses of the transformer windings and the non-sinusoidal excitation losses of the magnetic core, resulting in a deviation between the optimization results and the actual energy loss of the converter. Fourth, this scheme requires mode division and online calculation for each operating point to determine the shift ratio, and the online calculation burden is still relatively large.

[0005] In addition to the above-mentioned solutions, other efficiency optimization strategies have been disclosed in the prior art, mainly including the following categories: 1. Fixed-parameter or single-objective optimization control: In a frequency converter-based linearized control framework, the phase shift coefficient k is set to a constant, or the analytically optimal k value is obtained only for a single objective (such as minimizing reactive power or minimizing current stress). This results in the optimal k value under a single operating condition not being able to cover the entire operating range. For example, a small k value that helps reduce conduction losses under heavy load may cause a large number of switches to lose ZVS under light load, leading to a surge in switching losses; and vice versa. Fixed-parameter strategies cannot adaptively adjust according to changes in load and input voltage, resulting in limited efficiency optimization effects.

[0006] 2. Multi-mode hybrid modulation strategy: This approach artificially divides the operating range into multiple zones based on the load or voltage range, employing different modulation modes or k-values ​​in each zone, and achieving approximate optimization through mode switching. However, the selection of mode boundaries relies on experience or simplified theoretical analysis, making it difficult to guarantee global optimality. Mode switching may induce transient fluctuations in current or voltage, increasing control complexity. Furthermore, this method is essentially still a piecewise constant optimization, failing to achieve smooth and continuous efficiency optimization.

[0007] 3. Real-time optimization based on complex online algorithms: The loss model is run in real time in the controller, and the optimal control parameters are searched online. Accurate loss models (especially those involving switching transients and core nonlinearity) are computationally intensive, and online iterative optimization is time-consuming, requiring high processor performance, making it difficult to complete in real time in low-cost micro-inverters. At the same time, the convergence and stability of the online search are also difficult to guarantee.

[0008] Therefore, how to minimize the total loss of a DAB microinverter under all operating conditions within a wide range of photovoltaic voltage and load, while simultaneously considering the real-time constraints of the controller and the requirement for low-cost implementation, has become a pressing technical challenge in this field. Summary of the Invention

[0009] In view of this, the purpose of this invention is to provide a DAB micro-inverter efficiency optimization control method and system based on a loss model, which solves the problem that the efficiency of dual active bridge micro-inverters is difficult to achieve globally optimal operation over a wide range of photovoltaic voltages and loads. Specifically, the following technical problems need to be solved: how to accurately determine the optimal phase shift coefficient that minimizes total losses; how to achieve adaptive and smooth adjustment of the phase shift coefficient across the entire operating range; and how to meet the real-time constraints of the controller while ensuring the optimization effect and avoiding excessive online calculation burden.

[0010] This method solves the problems of existing technologies being unable to adaptively optimize total losses under all operating conditions and having a large online computational burden by establishing a variable frequency linearized control model and a loss model containing multiple losses, using the phase shift coefficient as a variable for offline global optimization and generating a lookup table.

[0011] To achieve the above objectives, the present invention provides the following technical solution: The efficiency optimization control method for dual active bridge microinverters based on a loss model provided by this invention includes the following steps: Step S1: Establish a system model for frequency conversion linearization control, where the phase shift coefficient k is defined as the ratio of the inner phase shift angle to the outer phase shift angle, and the switching frequency is made to vary with the inner phase shift angle so that the output current has a linear relationship with the inner phase shift angle; Step S2: Establish a loss model, which includes conduction loss, switching loss, transformer winding loss, and transformer core loss; Step S3: With minimizing the average total loss of the power frequency cycle as the optimization objective and the phase shift coefficient k as the optimization variable, an optimization algorithm is used to perform global optimization for each operating point in the entire operating range to obtain the optimal phase shift coefficient corresponding to each operating point, and an optimal phase shift coefficient lookup table covering the entire operating range is generated offline. The lookup table is indexed by photovoltaic voltage and output power. Step S4: When the dual active bridge microinverter is running online, the current photovoltaic voltage and the expected output power are collected in real time. The current photovoltaic voltage and the expected output power are used as indexes to query the lookup table to obtain the corresponding optimal phase shift coefficient. Step S5: Based on the optimal phase shift coefficient, an extended phase shift modulation strategy is adopted to generate pulse width modulation signals for controlling each switching transistor in the dual active bridge microinverter.

[0012] Furthermore, the frequency conversion linearization control satisfies the following linearization control law:

[0013] in, For output current, The transformer turns ratio Photovoltaic voltage, This is the inward phase angle. To calculate the equivalent leakage inductance on the secondary side, This is the maximum switching frequency allowed by the hardware.

[0014] Furthermore, in step S2, the switching loss model divides the switching process into a hard switching region, an incomplete zero-voltage turn-on region, and a complete zero-voltage turn-on region based on the magnitude and direction of the bridge arm current, and establishes an analytical expression for the energy loss of a single switch for each region.

[0015] Furthermore, in step S2, the transformer core loss model is designed for the trapezoidal wave characteristics of the transformer flux in a dual active bridge microinverter. An improved generalized Steinmetz formula is used to model the core loss under trapezoidal wave excitation, and constant terms related to core volume and material are eliminated by a reference operating condition normalization method.

[0016] Furthermore, the optimization algorithm in step S3 is a genetic algorithm; the genetic algorithm includes: using real number encoding to directly represent each individual as a real value of the phase shift coefficient k, uniformly and randomly generating an initial population within the feasible interval, using the sum of the reciprocal of the average total loss of the power frequency cycle and the frequency over-limit penalty term as the fitness function, generating offspring through tournament selection, simulated binary crossover and polynomial mutation, and using an elite retention strategy to ensure that the best individuals are not lost.

[0017] Furthermore, the full operating range in step S3 includes the entire operating range of photovoltaic voltage and the entire operating range of output power; the optimal phase shift coefficient lookup table is a two-dimensional lookup table; in step S4, if there is no index point in the lookup table that completely corresponds to the current photovoltaic voltage and the desired output power, the optimal phase shift coefficient corresponding to the current operating condition is calculated by a two-dimensional linear interpolation method.

[0018] Furthermore, the phase shift coefficient k in step S1 satisfies: Outward phase angle:

[0019] The switching frequency satisfy:

[0020] in, This indicates the switching frequency corresponding to the forward phase shift angle φ; Inward phase angle The outward phase angle is defined as the ratio of the time that the first and fourth switches in the primary-side full-bridge conduct together within half a cycle to the switching cycle. Defined as the ratio of the phase difference between the switching time of the secondary-side switch and the phase difference between the third and fourth switches on the primary side to the switching period; Represents the phase shift coefficient; This indicates the maximum switching frequency allowed by the hardware.

[0021] Furthermore, the transformer winding loss model in step S2 adopts an engineering modeling strategy based on frequency sweep measurement: first, frequency sweep test is performed on the primary and secondary windings of the transformer within the operating frequency range to obtain the measured AC resistance values ​​corresponding to discrete frequency points, and then a continuous functional relationship between the AC resistance factor and the switching frequency is established by using quadratic polynomial fitting, thereby calculating the total winding loss.

[0022] This invention provides an efficiency optimization control system for a dual active bridge microinverter based on a loss model, comprising: The main circuit of the half-bridge dual active bridge converter includes a photovoltaic-side full-bridge circuit, a high-frequency isolation transformer, an equivalent leakage inductance, a secondary half-bridge circuit, and a filter inductor. A sampling circuit, connected to the main circuit, is used to collect the current photovoltaic voltage and the desired output power; The memory contains a pre-generated offline lookup table of the optimal phase shift coefficient covering all operating conditions, indexed by photovoltaic voltage and output power. The optimal phase shift coefficient is a control parameter that minimizes the total losses of the converter. The controller is connected to the sampling circuit, the memory, and the main circuit respectively. The controller is configured to: obtain the optimal phase shift coefficient by querying a lookup table in the memory based on the current photovoltaic voltage and the expected output power collected by the sampling circuit, and generate a pulse width modulation signal for controlling each switch in the main circuit based on the optimal phase shift coefficient using an extended phase shift modulation strategy.

[0023] Furthermore, the optimal phase shift coefficient lookup table is generated through the following offline process: Establish a system model for variable frequency linearization control, where the phase shift coefficient is defined. and set the switching frequency This makes the output current correlate with the inner phase shift angle. The relationship is linear; Construct an accurate loss model that includes a conduction loss model, a switching loss model, a transformer winding loss model, and a transformer core loss model; The optimization objective is to minimize the average total loss during the power frequency cycle, with the phase shift coefficient as the key factor. To optimize the variables, a genetic algorithm is used to perform global optimization for each operating point within the entire operating range, obtaining the optimal phase shift coefficient for each operating point, thereby generating the lookup table.

[0024] The beneficial effects of this invention are as follows: This invention provides a method and system for efficiency optimization control of dual active bridge micro-inverters based on a loss model. The method first establishes a system model for frequency conversion linearization control, defining the phase shift coefficient as the ratio of the inner and outer phase shift angles and ensuring a linear relationship between the output current and the inner phase shift angle. Then, it establishes an accurate loss model including conduction losses, switching losses, transformer winding losses, and core losses. With the goal of minimizing the average total loss over the power frequency cycle, and using the phase shift coefficient as the optimization variable, an optimization algorithm is used to perform offline global optimization across all operating conditions, generating an optimal phase shift coefficient lookup table covering all operating conditions. During online operation, photovoltaic voltage and output power are collected in real time, the optimal phase shift coefficient is obtained by querying the lookup table, and extended phase shift modulation is used to generate a pulse width modulation signal. This invention achieves adaptive optimization of total losses across all operating conditions by combining offline global optimization with online table lookup. It has a low computational burden, is easy to implement in low-cost controllers, and significantly improves inverter operating efficiency. Compared with existing technologies, this invention has the following advantages: First, the optimization objective is more direct, and the results more closely reflect actual losses. This invention directly aims to minimize the total average loss, including conduction loss, switching loss, winding loss, and core loss. It eliminates the need for soft switching as a pre-constraint, allowing the optimization algorithm to automatically balance the benefits of soft switching with the costs of conduction losses. Compared to existing technologies that presuppose "achieving zero-voltage conduction across the entire power range," the optimization results of this invention more closely reflect the actual energy loss of the converter, avoiding additional conduction losses caused by forcibly satisfying soft-switching conditions.

[0025] Second, the optimization variables are simpler, and the optimization method is more efficient. This invention transforms the loss optimization problem into a global optimization problem with a single phase shift coefficient k as the variable. A genetic algorithm is used for offline global optimization, generating a two-dimensional optimal phase shift coefficient lookup table covering all operating conditions. Compared with existing technologies that rely on piecewise analytical solutions involving multiple phase shifts, this invention avoids complex mode division and online analytical calculations, achieving smooth and continuous optimality of the phase shift coefficient across all operating conditions.

[0026] Third, the loss model is more complete and the optimization accuracy is higher. This invention not only establishes an accurate switching loss model that distinguishes between hard switching, incomplete ZVS, and complete ZVS states, but also systematically incorporates the high-frequency eddy current loss of the transformer winding and the trapezoidal wave excitation loss of the magnetic core. The loss model has a wider coverage and higher accuracy, and can fully combine the measured data of the device and the characteristics of the actual magnetic components, ensuring a high degree of consistency between the optimization results and the actual physical system.

[0027] Fourth, the control implementation is more lightweight and has stronger engineering practicality. This invention completes all complex global optimization calculations offline. During online operation, only simple two-dimensional table lookup and interpolation calculations are needed based on the current photovoltaic voltage and output power, without complex iterative calculations, thus not increasing the real-time computing burden of the controller. Compared with existing technologies that require online mode division and parameter calculation, this invention is more suitable for widespread application in micro-inverter controllers with limited processor resources and cost sensitivity.

[0028] Fifth, the method exhibits full-condition adaptiveness, resulting in a significant efficiency improvement. Under heavy load, this method automatically reduces the phase shift coefficient k to decrease the effective current and reactive circulating current, significantly reducing conduction losses. Under light load, it automatically increases k to widen the zero-voltage switching range, significantly reducing switching losses and achieving a dynamic balance of losses across the entire power range. Simulation results show that the optimized efficiency at rated load is approximately 0.39 percentage points higher than the fixed parameter strategy, and the improvement is even more significant at light load, reaching 1.58 percentage points.

[0029] The above and other objects, advantages, and features of the present invention will be more fully set forth and demonstrated through the following detailed description of specific embodiments in conjunction with the accompanying drawings. Those skilled in the art, upon referring to the following detailed description and the accompanying drawings, will be able to better understand and realize the above advantages of the present invention. Other objects, features, and advantages of the present invention will become clearer after being described in detail in the detailed description section in conjunction with the accompanying drawings. Attached Figure Description

[0030] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the following drawings are provided for illustration.

[0031] Figure 1This is a topology diagram of a DAB microinverter; Figure 2 The drive signals and typical voltage and current waveforms for each switching transistor are shown. Figure 3 The waveforms of a general-purpose half-bridge switching unit and its switching process are shown in (a) circuit and (b) voltage and current waveforms. Figure 4 The waveforms are the primary voltage vp and magnetic flux density B of the transformer. Figure 5 A flowchart of the numerical computation of the objective function for a single candidate k value; Figure 6 Here is a flowchart of the genetic algorithm evolution process; Figure 7 Optimize the flowchart for all operating conditions; Figure 8 This is a control block diagram for online table lookup; Figure 9 The efficiency comparison curves before and after optimization are shown when Vpv is 35V. Detailed Implementation

[0032] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0033] Example 1

[0034] This embodiment provides a DAB microinverter efficiency optimization control method based on a loss model. This method is based on frequency conversion linearization control and generates an optimal phase shift coefficient lookup table through offline global optimization. During online control, only a simple table lookup is needed to maximize efficiency. The following detailed descriptions of each part are provided in conjunction with the accompanying drawings. The abbreviations used in this embodiment have the following meanings: Dual Active Bridge (DAB); Microinverter; Extended Phase Shift (EPS); Loss optimization; Genetic Algorithm (GA).

[0035] 1. Establishment of a system model based on variable frequency linearization control like Figure 1 As shown, Figure 1 This is a topology diagram of a DAB microinverter. In this embodiment, the DAB microinverter topology includes a photovoltaic side, a high-frequency transformer isolation section, and a grid side. The photovoltaic side consists of photovoltaic modules and DC filter capacitors. It consists of a primary-side full bridge and a secondary-side full bridge. The primary-side full bridge is composed of switching transistors. Composition. The high-frequency transformer isolation section includes a high-frequency isolation transformer. and equivalent leakage inductance High-frequency isolation transformer The number of turns of the primary side is The number of secondary turns is The turns ratio is Equivalent leakage inductance It is the equivalent inductance of the transformer leakage inductance referred to the secondary side, serving as the core component for power transmission. The grid side consists of a secondary half-bridge and a filter inductor. Composition. The secondary half-bridge contains two bidirectional switching units. and two voltage divider capacitors Each bidirectional switching unit consists of two sub-switches connected in reverse series.

[0036] like Figure 2 As shown, Figure 2 The drive signals and typical voltage and current waveforms of each switch are shown in this embodiment, which employs an extended phase-shift modulation strategy. The primary-side full-bridge switches... All are based on a fixed duty cycle. Driven by a square wave signal, the two switching transistors on the same bridge arm are... Complementary phase difference enables conduction. Switching transistor. The drive signal leads the switching transistor. Switch transistor and The ratio of the time of simultaneous conduction within half a cycle to the switching cycle is defined as the inner phase shift angle. On the grid side, when the grid voltage is in the positive half-cycle, the switching transistor... and With duty cycle The square wave drive and complementary conduction, while the switching transistor and Always on; when the mains voltage is in the negative half-cycle, the drive logic is reversed. The switching time of the secondary-side switch is relative to that of the primary-side switch. The ratio of the phase difference to the switching period is defined as the outward phase shift angle. .

[0037] During a complete switching cycle, the circuit experiences six operating modes. The rate of change of voltage and current across the leakage inductor is different in each mode, and together they determine the leakage inductor current i. s The waveform. By analyzing the circuit state under each operating mode and utilizing the half-cycle symmetry characteristic of the leakage inductance current, the expressions for the average transmission power and output current of the DAB micro inverter under steady state can be derived. The average transmission power is shown in equation (1): (1) in, This represents the average transmission power of the DAB microinverter. Indicates the transformer turns ratio; Indicates photovoltaic voltage; This represents the instantaneous value of the grid voltage; Indicates the high-frequency switching period; This represents the transformer equivalent leakage inductance referred to the secondary side; Indicates the inward phase shift angle; Indicates the outward phase angle; The expression for the output current is shown in equation (2): (2) Output current i o With control variables The relationship between φ and φ is complex and nonlinear, making direct application to closed-loop control extremely inconvenient. To simplify the control structure, a phase shift coefficient k is introduced, and a linear relationship is defined between the inner and outer phase shift angles: (3) Furthermore, to address the remaining nonlinearity issues, frequency conversion control is introduced based on extended phase-shift modulation, setting the switching frequency f... s With the inward shift phase angle It changes according to the following functional relationship: (4) In the formula, Represents the phase shift coefficient; This indicates the switching frequency corresponding to the forward phase shift angle φ; f smax The maximum switching frequency allowed by the hardware is determined by the switching speed of the power device and the design of the magnetic components.

[0038] Substituting equations (3) and (4) into the expression for the output current and simplifying, we obtain the simplified linearized control law: (5) in, For output current, Photovoltaic voltage, This is the inward phase angle. The maximum switching frequency allowed by the hardware. To calculate the equivalent leakage inductance on the secondary side, The transformer turns ratio is used; this control law makes the output current linearly related to the inner phase shift angle, thus simplifying the design of the grid-connected current tracking controller.

[0039] This expression makes the output current relative to... It exhibits a linear response, greatly simplifying the design of grid-connected current tracking controllers. Within this control framework, the maximum switching frequency f... smaxSince the hardware constraints are fixed in advance, the phase shift coefficient k becomes the only key parameter that can be adjusted in real time. The selection of the value of k directly affects the reactive power, the leakage inductance current at the switching moment, the degree of soft switching implementation, and the effective value of the current, thus fundamentally determining the loss distribution of the converter. This is the fundamental reason why this invention chooses k as the optimization variable.

[0040] 2. Accurate loss modeling for full-condition optimization To achieve optimization aimed at minimizing total loss, a loss analytical model that accurately reflects the physical process must be established. The instantaneous total loss p within one power frequency cycle... loss (t,k) consists of four parts: conduction loss, switching loss, winding loss, and core loss. (6) in, This represents the total instantaneous loss at time t and with a phase shift coefficient of k within a power frequency cycle. This represents the power device conduction loss at time t and phase shift coefficient k within the power frequency cycle. This represents the switching loss at time t and phase shift coefficient k within the power frequency cycle. This represents the high-frequency loss of the transformer winding at time t and phase shift coefficient k within the power frequency cycle. This represents the transformer core loss at time t and phase shift coefficient k within the power frequency cycle. (1) Conduction loss model Conduction losses depend on the effective value of the current during conduction and the on-resistance. In a full-bridge primary circuit, two switches are always on, and the current i flowing through the transformer primary side is... p Two switching transistors are also connected in series on the conduction path of the secondary half-bridge switch arm during any half-cycle, through which the leakage inductance current i flows. s Therefore, the total conduction loss is: (7) In the formula, R DS(on)_p and R DS(on)_s The on-resistance of the primary and secondary switching transistors can be obtained from the device datasheet.

[0041] in, Indicates the total conduction loss; This represents the total conduction loss of the primary-side switch on the photovoltaic side; This represents the total conduction loss of the secondary-side switching transistor on the power grid side; This represents the effective value of the primary current of the transformer; This represents the drain-source on-resistance of the primary-side switching transistor; This indicates the effective value of the secondary leakage inductance current; This indicates the on-resistance of the secondary-side switching transistor; The leakage inductance current at each switching moment can be accurately calculated by integration from the analytical value of the leakage inductance current and the piecewise linear waveform. The calculation formula is shown below: (8) in, This indicates the on-resistance of the secondary-side switching transistor; (2) Switching loss model like Figure 3 As shown, Figure 3 A general half-bridge switching unit for switching loss analysis and its switching process waveforms are presented. Figure 3 (a) in the diagram represents a general-purpose half-bridge switching unit circuit. Figure 3 (b) shows the voltage and current waveforms during the switching process, used to define the three operating regions: hard switching, incomplete ZVS, and complete ZVS. To establish an accurate switching loss model, this embodiment abstracts both the primary-side full-bridge and the secondary-side half-bridge into this general half-bridge unit for unified analysis.

[0042] Figure 3 (a) and v HB This indicates the input DC voltage of the general-purpose half-bridge unit, with its reference direction set to positive at the top and negative at the bottom; i HB This represents the bridge arm current flowing out from the midpoint of the half-bridge, with its reference direction being... Figure 3 The direction indicated by the arrow in (a) is the positive direction, meaning the flow from the midpoint of the bridge arm towards the external load or the lower pipe. Figure 3 In (b), t on This represents the duration of the overlap between the voltage drop across the transistor and the rise in channel current during the turn-on process (i.e., the turn-on transient time); t off This represents the duration of overlap between the voltage rise and the channel current drop during the turn-off process of the switching transistor (i.e., the turn-off transient time); t d This refers to the protection dead time set in the hardware control to prevent direct connection between the upper and lower bridge arms.

[0043] Based on the dead time t d Inner bridge arm current i HB The size and orientation of the switching process divide it into three working regions with clear physical meaning: Hard switching (HS) area: i H_B ≤0, the current direction is negative, the voltage across the lower transistor is still not discharged when the dead zone ends, the voltage and current overlap severely at the moment of turn-on, and the switching loss is the greatest.

[0044] Incomplete ZVS (iZVS) region: 0 H_B ≤2Q​oss / t d The current direction is positive but the amplitude is small, which is insufficient to completely remove the output capacitor charge within the dead time. The lower transistor turns on under the residual voltage, resulting in some overlap loss.

[0045] Full ZVS (ZVS) region: i H_B >2 Q oss / t d The current is large enough to completely discharge the voltage of the lower transistor to zero before the dead zone ends, achieving ideal zero-voltage turn-on, and the turn-on loss is negligible.

[0046] Among them, i H_B Q represents the bridge arm current flowing from the midpoint of the general-purpose half-bridge during the dead time; oss This indicates the amount of charge on the output capacitor of the switching transistor; For each region, this method establishes the single-switch energy loss E. sw The analytical expression for this. The total loss in a single switch operation is composed of the turn-off loss E. off , turn-on loss E on Dead zone loss E dt and reverse recovery loss E rr Composition. For GaN devices, reverse recovery loss is negligible, but the third-quadrant conduction loss generated by the reverse conduction channel during the dead time must be taken into account. The loss expressions for each region are detailed below: The switching loss in the hard-switching region can be modeled as follows: (9) In the formula, V SD For reverse conduction voltage drop, t on Q is the turn-on voltage drop time of the switching transistor. rr The reverse recovery charge of the body diode; This indicates the turn-off loss in the hard-switching region; This indicates the dead-zone loss in the hard-switching region; This indicates the turn-on loss in the hard-switching region; This indicates the reverse recovery loss in the hard-switching region; This indicates the reverse conduction voltage drop of the power switch transistor; Indicates the bridge arm current; This indicates the input voltage of the half-bridge circuit; This represents the voltage across the parasitic capacitance of the switching transistor; This represents the parasitic output capacitance of the switching transistor, the value of which is a nonlinear function of the voltage v across its terminals; This represents the reverse recovery charge of the body diode.

[0047] The switching loss in the incomplete ZVS region can be modeled as follows: (10) In the formula, Δv is the residual voltage remaining on the lower switching transistor at the end of the dead zone of the incomplete ZVS region. This indicates the turn-off loss in the incomplete ZVS region; This indicates the dead zone loss in the incomplete ZVS region; This indicates the turn-on loss of an incomplete ZVS region; This represents the reverse recovery loss in the incomplete ZVS region; The residual voltage Δv can be solved using the charge conservation equation, which is shown below: (11) The switching loss in the fully ZVS region can be modeled as follows: (12) in, This represents the turn-off loss in the complete ZVS region; This represents the dead zone loss in a fully ZVS region; This represents the turn-on loss of a fully ZVS region; This represents the reverse recovery loss in the incomplete ZVS region; Finally, the average switching loss power over one high-frequency switching cycle is: (13) in, This represents the average switching loss power over a complete high-frequency switching cycle. This represents the total energy loss of a single switching action within a switching cycle. Indicates the current high-frequency switching frequency; (3) Transformer winding loss model Winding losses are highly sensitive to switching frequency; at high frequencies, the skin effect and proximity effect can cause AC resistance to be several times greater than DC resistance. This invention employs an engineering modeling strategy based on frequency sweep measurement. First, an impedance analyzer is used to perform frequency sweep tests on the primary and secondary windings of the transformer within the operating frequency range, obtaining a series of measured AC resistance values ​​R corresponding to discrete frequency points. ac (f s_k The AC resistance factor at each frequency point is calculated using the following formula: F R (f s )=R ac (f s ) / R dc .

[0048] Among them, F R (f s ) indicates at the current switching frequency fs AC resistance factor under; R ac (f s R represents the measured value of AC resistance at the current switching frequency. dc This indicates the DC resistance of the transformer windings; Finally, a quadratic polynomial fitting method was used to establish F. R with f s Continuous functions: (14) Where a and b are fitting coefficients, f s0 For reference frequency; Based on this, the total loss of the transformer winding is: (15) in, express; This represents the AC resistance factor of the primary winding of a transformer. This represents the DC resistance of the primary winding of the transformer; Indicates the AC resistance factor of the secondary winding of a transformer; This represents the DC resistance of the secondary winding of the transformer; (4) Core loss model Core loss is the second most critical loss component in high-frequency transformers after winding loss, and its accurate modeling faces two major challenges: non-sinusoidal flux waveform and wide-range variation in switching frequency. In the DAB micro-inverter of this invention, the high-frequency square wave voltage output from the primary-side full-bridge is applied across the transformer's magnetizing inductance after internal phase-shift modulation, causing the magnetic flux density B(t) in the core to exhibit a typical trapezoidal waveform profile, such as... Figure 4 As shown, Figure 4 The primary voltage of the transformer is v p The graph shows the magnetic flux density B waveform, used to illustrate that under extended phase-shift modulation, the voltage applied to the transformer is a high-frequency square wave, resulting in a typical trapezoidal waveform profile for the magnetic flux density in the core. During the positive half-cycle (0~T... s Within / 2), the voltage waveform includes a forward excitation phase and a flat-top phase. The time period for the forward excitation phase is... This corresponds to the linear rising edge of the magnetic flux density from the negative peak to the positive peak in the diagram; when the voltage becomes zero, it enters the flat-top plateau region, which lasts for approximately [duration missing]. Subsequently, during the negative half-cycle, the magnetic flux density symmetrically experiences a linear falling edge and a negative plateau region.

[0049] Peak value of magnetic flux density B m It is determined by the integral of the applied voltage over time. During half a switching cycle, the excitation voltage +V pv The duration of action is According to Faraday's law of electromagnetic induction, we can deduce that: (16) Where, N p A is the number of turns in the primary winding. e This represents the effective cross-sectional area of ​​the magnetic core.

[0050] From the above formula, we can see that the inward phase angle Directly controlling the flux swing amplitude, and thus affecting the core loss. During the flux rise and fall phases, the absolute value of the rate of change of flux is: (17) During the plateau phase, the rate of change is zero, and no loss occurs. Therefore, core loss only occurs during the rising and falling edges. For this non-sinusoidal flux excitation with a flat-top segment and variable frequency, the traditional Steinmetz formula based on the sinusoidal assumption is no longer applicable. This invention uses the improved generalized Steinmetz formula (iGSE) to model core loss. Its core idea is to express the instantaneous loss density as a joint function of magnetic susceptibility and flux amplitude: (18) In the formula, C i These are constants derived from the standard Steinmetz coefficients, where α and β are material loss exponents. This represents the core loss per unit volume of the transformer. Indicates the high-frequency switching period; This represents the time variable for integration.

[0051] Substituting B(t) and dB(t) / dt of each stage of the trapezoidal wave into equation (18), and utilizing the symmetry of the positive and negative half-cycles, the integral is performed piecewise along the entire switching cycle. Since the rate of change of the plateau segment is zero, the integral only has contributions from the rising and falling edges. Through analytical derivation, the analytical expression for the core loss per unit volume of the transformer is obtained: (19) To avoid the effective volume V of the magnetic core in practical applications e To address the additional errors introduced by inaccurate measurements or temperature effects, this invention introduces a reference operating condition normalization method. A known operating condition is selected. The total core loss of the transformer under this operating condition can be accurately calibrated through experiments or finite element simulation. By taking the ratio of equation (19) under arbitrary and reference operating conditions and eliminating the constant terms related to volume and material, the total core loss under arbitrary operating conditions is obtained as follows: (20) in, This represents the total core loss of the transformer under any operating condition. This represents the total core loss calibrated through simulation or experiment under reference operating conditions. Indicates the inward phase shift angle under the reference operating condition; Indicates the switching frequency under reference operating conditions; Indicates the photovoltaic voltage under reference operating conditions; Equation (13) is the core loss model ultimately adopted in this invention. It incorporates the magnetic flux trapezoidal wave effect, frequency conversion operation characteristics, and non-sinusoidal excitation effects into a unified analytical framework. Simultaneously, it uses measured values ​​from reference operating conditions for self-calibration, eliminating the need to obtain the precise geometric dimensions and material microconstants of the core, thus improving the model's engineering applicability and accuracy. This model directly uses V... pv , and f s As input, it is seamlessly embedded into the aforementioned numerical calculation process of the objective function.

[0052] 3. Offline Global Optimization and Online Table Lookup Control Based on Genetic Algorithm To comprehensively evaluate the energy loss over a complete power frequency cycle, the average total loss over the power frequency cycle is defined as the optimization objective function: (twenty one) in, This represents the average total loss over a complete power frequency cycle. This represents the power frequency period, with a typical value of 20ms. Indicates the instantaneous total loss; The constraints of the optimization problem include: the range of k (0, 0.5] and the hardware limit f of the switching frequency. smin ≤f s (t)≤f smax In summary, the optimization problem can be fully expressed as: (twenty two) in, Indicates constraints; Indicates the minimum switching frequency allowed by the hardware; Indicates the maximum switching frequency allowed by the hardware; This represents the instantaneous switching frequency at time t within the power frequency cycle; Since the grid voltage and output current change continuously within the power frequency cycle, direct analytical integration is not possible. Therefore, this invention uses a numerical discretization method to calculate P. loss,avg .

[0053] like Figure 5 As shown, Figure 5 The flowchart shows the numerical calculation steps for the objective function under a single candidate k value. It illustrates the steps for obtaining the average total loss by discretizing the chemical frequency period, calculating the instantaneous loss at each discrete point, integrating, and finally obtaining the average total loss for a given operating condition and k value.

[0054] like Figure 5 As shown, for a given candidate phase shift coefficient The specific numerical calculation process is as follows: Step 1: Initialize discretization parameters.

[0055] One power frequency cycle (20ms) divided into equal parts In this embodiment, there are discrete sampling points. (That is, one sampling point every 1 electrical angle), corresponding to the time interval Set the current loop variable. and initialize the average loss accumulation value. and the maximum frequency limit .

[0056] Step 2: Iterate through the power frequency cycle.

[0057] For each sampling point Perform the following calculations: (2.1) Calculate the instantaneous values ​​of the grid voltage and output current: (twenty three) (twenty four) in, This is the effective value of the grid voltage. The power grid frequency is 50Hz. This is the effective value of the grid-connected current.

[0058] (2.2) Calculate the inner phase shift angle and instantaneous switching frequency at the current sampling time: Based on the frequency conversion linearization control law described in claim 1, the required inner phase shift angle at that moment is calculated in reverse. : (25) Then, substituting the values ​​into the frequency conversion function, the instantaneous switching frequency at that moment is calculated: (26) (2.3) Frequency over-limit judgment and maximum over-limit record: Determine the current switching frequency Does it meet hardware constraints? If the frequency meets the hardware constraints, no action is taken, and the process continues to the next step; otherwise, the maximum over-limit value is updated according to the formula: (27) (2.4) Calculate the instantaneous total loss using the loss model: Substitute all the variables calculated in steps (2.1) and (2.2) into the conduction loss established in the third part. Switching losses Winding losses and core loss In the calculation model, the instantaneous total loss at that sampling moment is calculated: (28) (2.5) Cumulative average loss: The instantaneous losses are accumulated to the average value for calculation, that is: (29) Step 3: Numerical integration and result output.

[0059] Determine if all have been traversed Each sampling point, i.e. Is it less than or equal to? If so, then let If not, return to step two and continue the loop; otherwise, calculate the average total loss per power frequency cycle for the candidate k value: (30) Finally return and the maximum frequency limit This is used for fitness evaluation in subsequent genetic algorithms.

[0060] The evolutionary module of the genetic algorithm is responsible for searching for the optimal k value that minimizes the average total loss within the feasible region. For example... Figure 6 As shown, Figure 6 The complete evolutionary process of the genetic algorithm is presented, with the specific steps as follows: Step 1: Population initialization: Real number encoding is used, with each individual directly represented as a real value of k. N individuals are uniformly and randomly generated within the feasible interval (0, 0.5]. pop (The value range is usually 20 to 100, and 50 is preferred in this embodiment. This size is sufficient to maintain population diversity, while the computational cost is small) individuals, which constitute the initial population P0.

[0061] Step 2 - Fitness Evaluation: For each individual k in the population, call the objective function numerical calculation module ( Figure 5 ) to get p loss,avg and the maximum frequency limit Δf m The fitness function is defined as follows: (31) Where ε is a small constant to prevent division by zero, typically taking the value of... λ is the penalty coefficient, and its typical range is... In this embodiment, the preferred value is .

[0062] in, P represents the fitness function value of an individual in a genetic algorithm. loss,avg The aforementioned average total loss per power frequency cycle This function enables individuals with low loss and who meet frequency constraints to achieve higher fitness.

[0063] Step 3 - Selection Operation: A tournament selection strategy is used. Each time, N are randomly selected from the population. tour Individuals are compared based on their fitness; the winner is placed in the mating pool. This process is repeated N times. pop Second-rate.

[0064] Step 4 - Crossover Operation: Randomly pair individuals in the mating pool with probability p. c (The value range is usually 0.6 to 0.95, and 0.85 is preferred in this embodiment. A higher crossover probability is beneficial for the rapid spread of superior genes in the population.) Perform simulated binary crossover (SBX) to generate offspring.

[0065] Step 5 - Mutation Operation: Mutate offspring individuals with probability p m (The value range is usually 0.01 to 0.1, and in this embodiment it is preferably 0.05. An appropriate mutation probability can prevent the algorithm from premature convergence and avoid degenerating the optimization into a random search.) Perform polynomial mutation to introduce new genetic diversity and prevent premature convergence.

[0066] Step 6 - Elite Preservation: Merge the parent and offspring populations, and select the N with the highest fitness. pop Each individual is treated as a new generation of the population, ensuring that the best individuals are not lost.

[0067] Step 7 - Termination Judgment: If the number of generations reaches G max (The value range is typically 50–200, and in this embodiment, 100 is preferred. For single-objective, univariate optimization problems, 100 generations are sufficient for the fitness function to converge to a stable value), or the optimal fitness change over multiple consecutive generations is less than a set tolerance (the tolerance threshold is set to...). If the iteration terminates, the optimal phase shift coefficient k under the current operating condition is output. opt And its corresponding minimum average total loss.

[0068] Figure 6 The process of iteratively searching for the optimal phase shift coefficient that minimizes the average total loss is presented through operations such as population initialization, fitness evaluation, selection, crossover, mutation, and elite retention.

[0069] like Figure 7As shown in the figure, the complete process of generating an offline two-dimensional lookup table of optimal phase shift coefficients covering all operating conditions is illustrated. "All operating conditions" here refers to all the operating states that a dual active bridge microinverter may face within a single grid-connected operation cycle, determined by the combination of photovoltaic input voltage and output power. In this embodiment, the operating range of the photovoltaic voltage is... The operating range of the output power is .

[0070] The specific offline optimization process for gridded scanning includes the following steps: Step 1: Set the discretization parameters for the operating conditions.

[0071] To balance optimization accuracy and offline computation efficiency, the working area is discretized into a grid. Photovoltaic voltage according to Discretize by step size, therefore The corresponding discrete sequence is It contains a total of 21 discrete voltage values; output power according to Discretize by step size, therefore The corresponding discrete sequence is It contains a total of 41 discrete power values.

[0072] Step 2: Full-condition mesh traversal.

[0073] All operating points are traversed using a nested loop: Set outer loop variable Traversal List, inner loop variable Traversal List. For each discrete operating point. The genetic algorithm evolution module described above is invoked to search for the optimal phase shift coefficient under this operating condition, with the goal of minimizing the average total loss per power frequency cycle. .

[0074] Step 3: Record the optimal parameters.

[0075] Each group The corresponding genetic algorithm optimization results And the The corresponding minimum average total loss is recorded in the data matrix.

[0076] Step 4: Save the lookup table.

[0077] After completing the traversal of all operating points, the above data matrix is ​​solidified and stored, forming a data matrix based on photovoltaic voltage. and output power Two-dimensional index, with optimal phase shift coefficient A two-dimensional lookup table covering the entire operating range for the values: .

[0078] In actual grid-connected operation, the above-mentioned global optimization process has been completed offline during the design phase, and the generated optimal phase shift coefficient lookup table is stored in the Flash memory of the digital controller.

[0079] like Figure 8 As shown, Figure 8 This is a block diagram of the online lookup table control for a DAB microinverter. The diagram illustrates the control logic and signal flow during online operation: the MPPT module provides the current photovoltaic voltage Vpv and the desired output power Po; the digital controller uses these two values ​​as indexes to look up the optimal phase shift coefficient from the table Table stored in the memory. kopt The optimal phase shift coefficient kopt is obtained; then the switching frequency fs and the inner phase shift angle are calculated sequentially. The phase angle φ is shifted outward and sent to the PWM modulation module, ultimately generating a drive signal to control the main circuit of the converter.

[0080] Perform the following operations during each control cycle: (1) The MPPT module obtains the current photovoltaic voltage V by sampling voltage and current. pv and expected output power P o ; (2) with (V) pv , P o ) is the index from the lookup table Table kopt (V pv ,P o The optimal phase shift coefficient k corresponding to the current operating condition is obtained through two-dimensional linear interpolation. opt ; (3) K opt Substitute into equation (4) to calculate the current switching frequency f. s ; (4) Based on the inner phase shift angle given by the current closed-loop controller Calculate the outward phase angle φ using equation (3); (5) , φ, f s The signal is fed into the PWM modulation module to generate a drive signal with precise dead-time compensation, which controls the on and off states of each power switch.

[0081] This solution completely transfers complex global optimization calculations to the offline design stage. Online control only needs to perform simple table lookups, interpolation, and algebraic operations, resulting in minimal computational overhead and strong engineering practicality. It can be easily implemented in cost-sensitive micro-inverter digital controllers.

[0082] 4. Optimize simulation results like Figure 9 As shown, Figure 9 To achieve photovoltaic voltage The simulation results curves comparing the efficiency before and after optimization are presented under operating conditions. This simulation verification was completed on the PLECS platform, with the following simulation conditions set: the effective value of the grid voltage is 220V / 50Hz; the turns ratio of the high-frequency transformer is set to 4, and the equivalent leakage inductance is 55μH; the maximum switching frequency is set to 80kHz. The system operates at a higher frequency (close to the upper limit) under light load and a lower frequency under heavy load. The switching frequency is controlled within the physical limits allowed by the hardware across the entire power range. This simulation model comprehensively considers non-ideal factors such as the parasitic parameters of power devices, the high-frequency characteristics of the transformer core and windings, etc., to accurately reflect the actual loss distribution of the converter.

[0083] To quantify the effect of optimization, in V pv At 35V, a fixed k of 0.35 was selected as the comparison benchmark. The average efficiency curves of the converter before and after optimization are shown below. Figure 9 As shown in the figure, the optimized efficiency is higher than the fixed k strategy across the entire power range. At the rated load P... o At 500W, the optimized efficiency reaches 97.13%, an improvement of approximately 0.39 percentage points compared to a fixed k; under light load P o At 100W, the improvement is even more significant, reaching 1.58 percentage points. The most obvious efficiency improvement is concentrated in the light load region, which reflects the optimization algorithm's proactive adjustment of k to balance switching and conduction losses. Table 1 shows the specific efficiency values ​​at key power points (e.g., 100W, 300W, 500W) before and after optimization.

[0084] Table 1. Comparison of critical power point efficiency before and after optimization at 35V photovoltaic voltage.

[0085] The above-described embodiments are merely preferred embodiments provided to fully illustrate the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are all within the scope of protection of the present invention. The scope of protection of the present invention is defined by the claims.

Claims

1. A method for efficiency optimization control of a dual active bridge microinverter based on a loss model, characterized in that, Includes the following steps: Step S1: Establish a system model for frequency conversion linearization control, where the phase shift coefficient k is defined as the ratio of the inner phase shift angle to the outer phase shift angle, and the switching frequency is made to vary with the inner phase shift angle so that the output current has a linear relationship with the inner phase shift angle; Step S2: Establish a loss model, which includes conduction loss, switching loss, transformer winding loss, and transformer core loss; Step S3: With minimizing the average total loss of the power frequency cycle as the optimization objective and the phase shift coefficient k as the optimization variable, an optimization algorithm is used to perform global optimization for each operating point in the entire operating range to obtain the optimal phase shift coefficient corresponding to each operating point, and an optimal phase shift coefficient lookup table covering the entire operating range is generated offline. The lookup table is indexed by photovoltaic voltage and output power. Step S4: When the dual active bridge microinverter is running online, the current photovoltaic voltage and the expected output power are collected in real time. The current photovoltaic voltage and the expected output power are used as indexes to query the lookup table to obtain the corresponding optimal phase shift coefficient. Step S5: Based on the optimal phase shift coefficient, an extended phase shift modulation strategy is adopted to generate pulse width modulation signals for controlling each switching transistor in the dual active bridge microinverter.

2. The efficiency optimization control method for a dual active bridge microinverter based on a loss model according to claim 1, characterized in that, The frequency conversion linearization control satisfies the following linearization control law: in, For output current, The transformer turns ratio Photovoltaic voltage, This is the inward phase angle. To calculate the equivalent leakage inductance on the secondary side, This is the maximum switching frequency allowed by the hardware.

3. The efficiency optimization control method for a dual active bridge microinverter based on a loss model according to claim 1, characterized in that, The switching loss model in step S2 divides the switching process into a hard switching region, an incomplete zero-voltage turn-on region, and a complete zero-voltage turn-on region based on the magnitude and direction of the bridge arm current, and establishes an analytical expression for the energy loss of a single switch for each region.

4. The efficiency optimization control method for a dual active bridge microinverter based on a loss model according to claim 1, characterized in that, The transformer core loss model in step S2 is designed for the trapezoidal wave characteristics of transformer flux in dual active bridge microinverters. It uses an improved generalized Steinmetz formula to model the core loss under trapezoidal wave excitation and eliminates constant terms related to core volume and material by using a reference operating condition normalization method.

5. The efficiency optimization control method for a dual active bridge microinverter based on a loss model according to claim 1, characterized in that, The optimization algorithm in step S3 is a genetic algorithm; the genetic algorithm includes: using real number encoding to directly represent each individual as a real value of the phase shift coefficient k, uniformly and randomly generating an initial population within the feasible interval, using the sum of the reciprocal of the average total loss of the power frequency cycle and the frequency over-limit penalty term as the fitness function, generating offspring through tournament selection, simulated binary crossover and polynomial mutation, and using an elite retention strategy to ensure that the best individuals are not lost.

6. The efficiency optimization control method for a dual active bridge microinverter based on a loss model according to claim 1, characterized in that, The full operating range in step S3 includes the entire operating range of photovoltaic voltage and the entire operating range of output power; the optimal phase shift coefficient lookup table is a two-dimensional lookup table; in step S4, if there is no index point in the lookup table that completely corresponds to the current photovoltaic voltage and the desired output power, the optimal phase shift coefficient corresponding to the current operating condition is calculated by a two-dimensional linear interpolation method.

7. The efficiency optimization control method for a dual active bridge microinverter based on a loss model according to claim 1, characterized in that, The phase shift coefficient k in step S1 satisfies: Outward phase angle: The switching frequency satisfy: in, This indicates the switching frequency corresponding to the forward phase shift angle φ; Inward phase angle The outward phase angle is defined as the ratio of the time that the first and fourth switches in the primary-side full-bridge conduct together within half a cycle to the switching cycle. Defined as the ratio of the phase difference between the switching time of the secondary-side switch and the phase difference between the third and fourth switches on the primary side to the switching period; Represents the phase shift coefficient; This indicates the maximum switching frequency allowed by the hardware.

8. The efficiency optimization control method for a dual active bridge microinverter based on a loss model according to claim 1, characterized in that, The transformer winding loss model in step S2 adopts an engineering modeling strategy based on frequency sweep measurement: first, frequency sweep test is performed on the primary and secondary windings of the transformer within the operating frequency range to obtain the measured AC resistance values ​​corresponding to discrete frequency points; then, a quadratic polynomial fitting is used to establish a continuous functional relationship between the AC resistance factor and the switching frequency, thereby calculating the total winding loss.

9. A dual active bridge microinverter efficiency optimization control system based on a loss model, characterized in that, include: The main circuit of the half-bridge dual active bridge converter includes a photovoltaic-side full-bridge circuit, a high-frequency isolation transformer, an equivalent leakage inductance, a secondary-side half-bridge circuit, and a filter inductor. A sampling circuit, connected to the main circuit, is used to collect the current photovoltaic voltage and the desired output power; The memory contains a pre-generated offline lookup table of the optimal phase shift coefficient covering all operating conditions, indexed by photovoltaic voltage and output power, wherein the optimal phase shift coefficient is a control parameter that minimizes the total losses of the converter; as well as The controller is connected to the sampling circuit, the memory, and the main circuit respectively. The controller is configured to: obtain the optimal phase shift coefficient by querying a lookup table in the memory based on the current photovoltaic voltage and the expected output power collected by the sampling circuit, and generate a pulse width modulation signal for controlling each switch in the main circuit based on the optimal phase shift coefficient using an extended phase shift modulation strategy.

10. The efficiency optimization control system for a dual active bridge microinverter based on a loss model according to claim 9, characterized in that, The optimal phase shift coefficient lookup table is generated through the following offline process: Establish a system model for variable frequency linearization control, where the phase shift coefficient is defined. and set the switching frequency This makes the output current correlate with the inner phase shift angle. The relationship is linear; Construct an accurate loss model that includes a conduction loss model, a switching loss model, a transformer winding loss model, and a transformer core loss model; The optimization objective is to minimize the average total loss during the power frequency cycle, with the phase shift coefficient as the key factor. To optimize the variables, a genetic algorithm is used to perform global optimization for each operating point within the entire operating range, obtaining the optimal phase shift coefficient for each operating point, thereby generating the lookup table.

Images