A deadbeat average model control method for current prediction of phase-shifted full-bridge converter

Abstract

The present invention discloses a current prediction deadbeat average model control method for a phase-shifted full-bridge converter. The control method combines oversampling technology to convert the current loop controlled object from a single-point sampling inductor current to the average value of the inductor current. The next beat control quantity is calculated by predicting the current value of the current cycle, thereby reducing the influence of the lagging one-beat control and improving the dynamic performance of the controller. The proposed method effectively improves the contradiction between the stability and dynamic performance of the DC / DC converter control system, thereby improving the overall stability and reliability of the modular multi-level digital power amplifier system. In addition, pulse feedforward control is introduced to address the problem of duty cycle loss of the phase-shifted full-bridge converter, which not only improves the steady-state tracking performance of the deadbeat control of the phase-shifted full-bridge converter, but also further improves the dynamic performance of the system and expands the matching range of the output inductor parameters.

Description

Technical Field

[0001] The invention belongs to the technical field of power electronic devices, and in particular relates to a current prediction deadbeat average model control method for a phase-shifted full-bridge converter. Background Art

[0002] The control system of the phase-shifted full-bridge converter can be a single-loop control system or a multi-loop control system according to the number of closed-loop control objects. The single-loop control system usually refers to the output voltage loop control. When the single-loop control is adopted, the system only takes the output voltage as feedback, and the system cannot accurately control the output current. In the multi-loop control system, multiple controlled objects can be selected according to different control requirements, which is suitable for occasions with high control performance requirements. In the phase-shifted full-bridge converter, there are usually requirements for both output voltage and output current. Therefore, the controlled objects include output voltage and inductor current, and dual-loop control is generally adopted. In order to ensure the constant voltage characteristic of the DC converter output, the voltage loop can be selected as the control outer loop, and the current loop as the control inner loop is responsible for tracking the given current. Introducing the current inner loop control on the basis of the voltage outer loop control not only improves the dynamic performance of the DC converter, but also controls the current flowing through the power device within the safe working range, greatly improving the stability and reliability of the DC converter system. Therefore, the current inner loop is a key factor affecting the performance of the control system.

[0003] The implementation method of the current inner loop using zero-beat control is to obtain the system status information at the current moment through sampling, combine the target value of the control object at the next sampling moment with the system circuit model, calculate the control signal required for the current cycle and apply the control without delay, so the dynamic performance is excellent.

[0004] The traditional deadbeat control method is based on single-point sampling control, and the current sampling point is usually set at the peak or valley point in the switching cycle for single-point sampling. As the power density of DC converters increases, especially in high-power applications, digital control systems are in a strong electromagnetic interference environment. The single-point sampling-based method is susceptible to common-mode and differential-mode interference caused by high-speed device switching actions, thereby affecting the system's dynamic control effect and output stability, and in extreme cases even causing system failure.

[0005] In order to improve the anti-interference ability, the digital control system needs to add a larger filtering link in the control loop, and then control the system according to the filtered sampling value. At the same time, there are system delays such as calculation and pulse update in the digital system, so the calculated control signal required for the current cycle can usually be updated in the next cycle. This method is called a one-beat output control method. Therefore, the addition of the filtering link and the one-beat control method make it difficult to bring into play the dynamic performance advantages of the traditional zero-beat control method. In addition, the duty cycle loss characteristic of the secondary side of the phase-shifted full-bridge topology also causes the average sampling point, peak sampling point, and valley sampling point time of the inductor current to drift with the input voltage and load current, so there is a problem of difficulty in selecting the sampling point time. Summary of the invention

[0006] In order to solve the shortcomings of the traditional deadbeat control in the above-mentioned phase-shifted full-bridge converter, the present invention uses multi-point sampling technology to propose a predictive deadbeat average model control method to improve the anti-interference performance and dynamic performance of the phase-shifted full-bridge inverter control system. The control method is used on the phase-shifted full-bridge converter, which is composed of a full-bridge inverter circuit, a resonant inductor, a high-frequency power transformer, a full-bridge rectifier circuit, and a L f and C f The output filter circuit of the composition is connected to form a double closed-loop control system, the voltage outer loop adopts PI control, and the current inner loop adopts a predictive deadbeat average model control method, and the predictive deadbeat average model control method is:

[0007] When the system runs in cycle k, the output filter inductor current is oversampled and averaged to obtain the average current value I of cycle k. avg (k);

[0008] When the system runs in the k+1 cycle, according to the average value of the inductor current I avg (k) and k+2 cycle current given value I set (k+2) calculates the control signal D of k+2 period k+2 , the calculation formula is:

[0009]

[0010] Among them, T s is the control period, L is the output filter inductance, D k is the k-cycle duty cycle, D k+1 is the duty cycle of the k+1 cycle obtained by the above formula in the k-1 cycle, K is the primary-secondary ratio of the high-frequency power transformer, U in is the input voltage of the phase-shifted full-bridge converter, U out is the output voltage of the phase-shifted full-bridge converter, I set(k+1) is the given current value of the k+1 cycle.

[0011] Since the phase-shifted full-bridge conversion topology has the phenomenon of secondary side duty cycle loss, if the lost duty cycle is not compensated, the deadbeat control has similar characteristics to proportional control, which makes it impossible to statically follow the steady-state value and will also affect the dynamic performance of the system. Therefore, the present invention also proposes a predictive deadbeat average model control method, which uses feedforward compensation to compensate for the duty cycle and effectively overcome the influence of the loss of the secondary side duty cycle on the control effect.

[0012] In the feedforward compensation, the secondary side lost duty cycle signal that needs to be compensated for k cycles can be expressed as:

[0013]

[0014] Among them, L r It is the sum of the resonant inductance and the leakage inductance of the high-frequency power transformer T.

[0015] The present invention proposes a predictive deadbeat average model control method for a phase-shifted full-bridge converter in combination with oversampling technology. The main features and advantages of the proposed control method are:

[0016] (1) By converting the traditional single-point current control into multi-point current average value control, the problem of single-point sampling being easily disturbed and causing the controller to lose stability is overcome. At the same time, a model-based prediction method is used to predict the next-beat control quantity, reducing the impact of the delayed one-beat control and improving the dynamic performance of the controller. The proposed method effectively improves the contradiction between the stability and dynamic performance of the phase-shifted full-bridge converter control system, thereby improving the overall stability and reliability of the modular multi-level digital power amplifier system.

[0017] (2) To solve the duty cycle loss problem of the phase-shifted full-bridge converter, pulse feedforward control is introduced. This not only improves the steady-state tracking performance of the phase-shifted full-bridge converter with zero-beat control, but also further improves the dynamic performance of the system and expands the matching range of the output inductor parameters.

[0018] (3) The predicted zero-beat average model control method is not only applicable to the phase-shifted full-bridge converter, but also applicable to all Buck-type DC converters by removing the duty cycle compensation link. It is a universal control method for improving stability and dynamic performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0020] Figure 1 This is the topology diagram of the phase-shifted full-bridge converter.

[0021] Figure 2 This is the voltage and current dual-loop control block diagram of the phase-shifted full-bridge converter.

[0022] Figure 3 Block diagram of the control principle of the predictive deadbeat average model for the phase-shifted full-bridge converter.

[0023] Figure 4 Schematic diagram of multi-point sampling.

[0024] Figure 5 Predictive deadbeat average model control block diagram for the phase-shifted full-bridge converter.

[0025] Figure 6 The waveforms of the average inductor current under interference for the two control methods are shown in Figure 2.

[0026] Figure 7 The average current step response of different model parameter matching before feedforward compensation, where (a) is K L =0.5, (b) is K L =0.75, (c) is K L =1.0, (d) is K L =2.0.

[0027] Figure 8 The average current step response of different model parameters after feedforward compensation, where (a) is K L =0.5, (b) is K L =0.75, (c) is K L =1.0, (d) is K L =2.0.

[0028] Fig. 9 The traditional deadbeat control inductor current waveform, where (a) is given at 30A and (b) is given at 60A.

[0029] Fig.10 To predict the inductor current waveform controlled by the deadbeat average model, (a) is given by 30A and (b) is given by 60A.

[0030] Fig.11 The traditional deadbeat control inductor current waveform under step setting, where (a) is 30A-60A step setting, and (b) is 60A-30A step setting.

[0031] Fig.12The predicted deadbeat average model controls the inductor current waveform under step reference, where (a) is a 30A-60A step reference and (b) is a 60A-30A step reference.

[0032] Fig.13 The inductor current waveform and output voltage waveform of the traditional deadbeat control method under dual closed loops, where (a) is a 32A load and (b) is a 78A load.

[0033] Fig.14 The inductor current waveform and output voltage waveform are predicted by the zero-beat average model control method under dual closed loops, where (a) is a 32A load and (b) is a 78A load. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0035] The phase-shifted full-bridge converter is composed of a full-bridge inverter circuit, a resonant inductor, a high-frequency power transformer, a full-bridge rectifier circuit, and a L f and C f The output filter circuit is connected and its topology is as follows: Figure 1 As shown, the power switches S1 and S3 form the leading bridge arm, S2 and S4 form the lagging bridge arm, and the control signals of the leading bridge arm and the lagging bridge arm are staggered at a certain angle to provide conditions for the zero voltage turn-on and turn-off of the switch device. The capacitor connected in parallel to the power switches S1 to S4 is the sum of the external resonant soft switch capacitor and the power device body capacitance, L r It is the sum of the resonant inductance and the leakage inductance of the high-frequency power transformer T. K is the primary-to-secondary ratio of the high-frequency power transformer. The secondary side of the transformer is composed of D R1 ~D R4 The rectifier filter circuit consists of L f and C f The output filter circuit is composed of

[0036] The phase-shifted full-bridge converter has 12 switching modes in one control cycle, and the following assumptions are made:

[0037] (1) All capacitors and inductors in the phase-shifted full-bridge converter adopt ideal models;

[0038] (2) Output filter inductor L f >>L r / K 2 ;

[0039] (3) Ignore the voltage fluctuation across the output capacitor during a switching cycle and assume that the voltage remains unchanged;

[0040] (4) Ignore the DC bias effect of the transformer.

[0041] The control system of the phase-shifted full-bridge converter adopts a voltage-current dual-loop control system, and the controlled objects are the output voltage and the inductor current. In order to ensure the constant voltage characteristic of the phase-shifted full-bridge converter output, the voltage loop can be selected as the control outer loop, and the current loop as the control inner loop is responsible for tracking the given current. The control block diagram of the system is shown in the figure below: Figure 2 As shown in the figure, G Cu (z), G Ci (z) is the controller transfer function, Z OH is the zero-order holder in the digital system, G id (z) is the transfer function of the duty cycle signal d to the inductor current, K i is the output voltage feedback loop gain, G ui (z) is the transfer function of the inductor current to the output voltage.

[0042] The block diagram of the control principle of the phase-shifted full-bridge converter predictive deadbeat average model is as follows: Figure 3 As shown, the voltage loop adopts PI control, and the current loop adopts the predictive deadbeat average model control method.

[0043] The predicted deadbeat average model control method is:

[0044] The output filter inductor current is oversampled and averaged to obtain the k-cycle current average value I avg (k);

[0045] When the system runs in the k+1 cycle, according to the average value of the inductor current I avg (k) and k+2 cycle current given value I set (k+2) calculates the control signal D of k+2 period k+2 , the calculation formula is:

[0046]

[0047] Among them, T s is the control period, L is the output filter inductance, D k is the k-cycle duty cycle, D k+1 is the duty cycle of the k+1 cycle obtained by the above formula in the k-1 cycle, K is the primary-secondary ratio of the high-frequency power transformer, U in is the input voltage of the phase-shifted full-bridge converter, U out is the output voltage of the phase-shifted full-bridge converter, I set (k+1) is the given current value of the k+1 cycle.

[0048] The derivation process of the predicted deadbeat average model control method is as follows:

[0049] (1) Deadbeat control average model

[0050] like Figure 4 As shown, we can start by establishing a deadbeat control model with multi-point sampling within a control cycle. Assume that in a control cycle T s There are n sampling points in the control cycle, and the input voltage and output voltage can be considered unchanged in one control cycle, and the root can be obtained:

[0051]

[0052] Among them, i j (k) is the sampling current at the jth point in the k control cycle, i j (k) and i j (k+1) The time interval is T s , d j is the duty cycle between sampling point j of the k control cycle and sampling point j of the k+1 control cycle. If the above formula is used alone for deadbeat control, the overall calculation amount will increase by n times, and the control cycle T s Also shorten to T s / n, the general controller is limited by the calculation speed and cannot meet the requirements. Let's add up the expressions in the above formula one by one:

[0053]

[0054] definition

[0055]

[0056] When the sampling frequency is much higher than twice the signal bandwidth, the sampling points are sufficient, which means oversampling technology. avg (k) can be considered as the average value of the inductor current.

[0057] Combined with formula (2), the expression on the left side of formula (2.16) can be transformed into:

[0058]

[0059] It can be seen that after multi-point sampling, the left side of equation (3) has been converted into the expression form of the average value of the inductor current.

[0060] The right side of formula (3) is difficult to solve in the discrete domain. When there are many sampling points, it can be equivalent to solving it in the continuous domain. Then the right side of formula (3) can be expressed as:

[0061]

[0062] Where T = Ts / n, T represents the sampling period, and defines D k is the k-cycle duty cycle, D k+1 is the duty cycle of the k+1 cycle. If we can get d j The expression of can be used to solve the integral of equation (7), and the duty cycle d can be calculated by different situations. j The expressions of can be obtained:

[0063]

[0064] The above formula obtains the relationship between the average inductor current and the duty cycle, which can be used as the basis for the zero-beat average model control.

[0065] According to formula (8), the average current I of the k+1 period is calculated. avg (k+1) prediction. Order:

[0066]

[0067] The difference equation of the deadbeat average model control can be obtained:

[0068]

[0069] If it is necessary to achieve an average current of a given value I in the k+1 period set The error-free tracking of (k+1) can be:

[0070]

[0071] According to equations (10) and (11), the deadbeat average model control method of the inductor current can be obtained as follows:

[0072]

[0073] It is not difficult to find from the above formula that the control object of the deadbeat average model control method is transformed from the traditional single-point sampling quantity to the average value of multi-point sampling quantities, which is conducive to improving the anti-interference performance of the digital control system. And from (8), it can be seen that multi-point sampling control only needs to oversample a single current quantity in one control cycle and calculate the average value, which greatly reduces the calculation amount of the algorithm.

[0074] (2) Predictive Control

[0075] When the system runs in the k+1 cycle, the oversampling process has not yet ended, so the average current value of the k+1 cycle cannot be obtained. Therefore, the predictive control is introduced to predict the average current of the k+1 cycle. At the same time, the equation (12) can be recursively deduced to obtain:

[0076]

[0077] The two equations in equation (13) are added and eliminated. The difference equation is:

[0078]

[0079] From equation (14), it can be seen that the current prediction of the k+2 cycle can be obtained through the state of the k cycle and the duty cycle of the k+1 cycle.

[0080] If we need to calculate the control signal D for k+2 cycles k+2 , to achieve the average current for a given value I set The error-free tracking of (k+2) can be obtained by:

[0081]

[0082] By combining equation (14) and equation (15), we can get:

[0083]

[0084] The above formula is the method of predicting the deadbeat average model control. Through formula (16), the k+2 cycle duty cycle control signal D can be obtained. k+2 , where I avg (k) is the average inductor current in the previous cycle, D k is the duty cycle of the previous cycle, D k+1 The duty cycle of the k+1 current cycle is obtained by predictive control in the k-1 cycle. Considering the loss of the secondary duty cycle of the phase-shifted full bridge, the above-mentioned duty cycle signals are all valid duty cycle signals. The next beat control quantity is calculated by predicting the current value of the current cycle. If the current prediction is accurate, it can theoretically offset the influence of the one-beat control method of the traditional digital controller.

[0085] (3) Controller design

[0086] Minimum beat control requires that the digital system ends the response process with the least finite beats under typical input and has no steady-state error at the sampling time, which is the basis of deadbeat control. The deadbeat controller is designed based on the minimum beat control theory.

[0087] According to the principle of closed-loop digital system feasibility, the system response cannot be simultaneous with or ahead of the input signal, and the system output must go through at least one control cycle to end the response process. Therefore, the closed-loop transfer function of the system contains at least one pure lag link. In digital control systems, due to the existence of sampling, calculation and pulse transmission delays, a one-beat lag control is used, and the system will contain another pure lag link z-1. Therefore, the closed-loop transfer function Φ(z) of the system can be expressed as:

[0088] Φ(z)=z-2 ·N(z) (17)

[0089] Where N(z) is the number not including (1-z -1 ) factor z -1 Polynomial.

[0090] It can be seen from the above formula that when the traditional minimum beat design principle is adopted, the digital closed-loop control system must go through at least two sampling cycles to complete the response process.

[0091] Aiming at the problem of one-beat delayed output of digital system, a design method of transfer function of deadbeat controller is proposed. Taking the step response of the system as an example, the transfer function of the input step signal is:

[0092]

[0093] The closed-loop transfer function of the system is designed according to the principle of minimum beats. In order to make the designed digital controller as simple as possible and the system transition process as fast as possible, N(z) = 1 is usually taken, then:

[0094]

[0095] Among them, Φ e (z) is the system error transfer function. After adopting the predictive deadbeat control method, equation (14) can be written as:

[0096]

[0097] The transfer function of the controlled object is:

[0098]

[0099] The transfer function of the deadbeat controller is:

[0100]

[0101] Therefore, the deadbeat controller under step input can be designed according to equation (22).

[0102] According to equations (18) and (19), the ripple effect of the system at non-sampling time is analyzed. The input sequence of the digital controller is:

[0103] E 1 (z) = R(z)·Φ e (z) = 1 + z -1 (twenty three)

[0104] The output sequence of the digital controller, that is, the input sequence of the controlled object is:

[0105]

[0106] It can be seen that the input sequence of the controlled object reaches relative stability and no fluctuation after a limited sampling period. Therefore, under step input, the system is a ripple-free minimum beat system.

[0107] In the above discussion, the average model of deadbeat control is derived by introducing the concept of multi-point sampling and the current loop controlled object is converted from single-point sampling inductor current to the average value of inductor current. In addition, in order to overcome the influence of one-beat control lag of the digital system, predictive control is introduced on the established average model of deadbeat control. According to the k+1 period system, the state I avg (k), D k and D k+1 The switching signals of k+2 cycles are calculated together, and finally the predictive deadbeat average model control method is obtained.

[0108] The above-mentioned predictive deadbeat average model control method is not only applicable to the phase-shifted full-bridge converter, but also to all other Buck-type DC converters. It is a universal control method for improving stability and dynamic performance.

[0109] Since the phase-shifted full-bridge conversion topology has the phenomenon of secondary duty cycle loss, if the lost duty cycle is not compensated, the zero-beat control has a characteristic similar to proportional control, which makes it impossible to statically follow the steady-state value, and will also affect the dynamic performance of the system. Therefore, another embodiment of the present invention adopts a feedforward compensation method to compensate for the duty cycle and effectively overcome the influence of the secondary duty cycle loss on the control effect. As shown in formula (25), the primary current I 1 ,I 3 Equivalent to the secondary current, the secondary lost duty cycle signal that needs to be compensated for k cycles can be expressed as:

[0110]

[0111] The final block diagram of the phase-shifted full-bridge converter predictive deadbeat average model control is as follows: Figure 5 However, in the actual implementation process, since the starting current and the ending current cannot be accurately collected, the average current of the inductor current is used instead. The improved formula is:

[0112]

[0113] In order to verify the correctness and effectiveness of the proposed method, a phase-shifted full-bridge converter simulation and test platform system was built. A phase-shifted full-bridge converter was built in Matlab / Simulink, and the circuit parameters are shown in Table 2.2.

[0114] Table 2.2 Phase-shifted full-bridge converter simulation parameters

[0115]

[0116]

[0117] In terms of control, the phase-shifted full-bridge voltage outer loop adopts PI control, and the current inner loop adopts traditional deadbeat control and the predicted deadbeat average model control proposed in this paper. During the simulation, since only the current loop performance is compared, the voltage outer loop does not work, the output is replaced by a voltage source, and the current loop input directly adopts the given value.

[0118] In order to verify the improvement of the proposed control algorithm in anti-interference performance, the predicted deadbeat average model control is compared with the traditional deadbeat control. At 25% rated current, the average current fluctuation range at different noise intensities is: traditional deadbeat control ±454.4%, ±375.7%, ±402.9%, predicted deadbeat control ±5.19%, ±8.99%, ±16.75%; at 50% rated current, the average current fluctuation range at different noise intensities is: traditional deadbeat control ±214.1%, ±219.2%, ±253.3%, predicted deadbeat control ±3.35%, ±7.58%, ±15.5%; at rated current, the average current fluctuation range at different noise intensities is: traditional deadbeat control ±151.0%, ±153.9%, ±161.3%, predicted deadbeat control ±4.88%, ±9.35%, ±16.14%. At the same time, under different current conditions, the two control methods injected high-frequency random noise with peak values ​​of 5%, 10%, and 20% of the rated current, and the current fluctuated to a certain extent at the interference moment and in the following control cycles. However, the anti-interference performance of the deadbeat control using the current average value is significantly better than that of the traditional deadbeat control.

[0119] Comparison of the predicted deadbeat average model control and the traditional deadbeat control inductor current at different load steps. During the simulation, the voltage outer loop is removed and a constant voltage source is used to maintain the stability of the output voltage, and the dynamic performance of the two control methods is compared by directly changing the input setting of the current inner loop. Under the rated current step of 25%-50%-25%, the average current overshoot of the traditional deadbeat control is 37.8% and 75.9%, and the adjustment time is 5.56e-5 and 1.96e-4; the average current overshoot of the predicted deadbeat control is 7.91% and 12.0%, and the adjustment time is 2.4e-6 and 2.8e-6; under the rated current step of 50%-75%-50%, the average current overshoot of the traditional deadbeat control is 26.0% and 42.3%, and the adjustment time is 3.32e-5 and 1. 96e-4; the predicted average current overshoots of deadbeat control are 4.9% and 7.0%, and the adjustment times are 1.6e-6 and 2.0e-6, respectively; under the rated current step of 10%-90%-10%, the average current overshoots of traditional deadbeat control are 38.4% and 74.4%, and the adjustment times are 4.36e-5 and 1.96e-4, respectively; the predicted average current overshoots of deadbeat control are 8.1% and 28.1%, and the adjustment times are 2.4e-6 and 1.57e-4, respectively.

[0120] From the above data sets, it can be seen that the current overshoot and adjustment time of the predicted deadbeat average model control method are greatly reduced, and the oscillation is small during the transition, so the dynamic performance of the predicted deadbeat average model control method is better than the traditional deadbeat control method.

[0121] Figure 6 The inductor current waveforms with and without duty cycle feedforward compensation are shown for different current step commands. Due to duty cycle loss in the phase-shifted full-bridge converter, there is a certain steady-state error in the uncompensated deadbeat control, so duty cycle loss compensation needs to be introduced. Before compensation, the actual output is 62.7A, which fails to follow the command value. After compensation, the output is 80.0A, which reaches the command value.

[0122] Under the rated current step of 25%-50%-25%, the average current overshoot before feedforward compensation is 19.3% and 38.5%, the adjustment time is 3.6e-6 and 5.6e-6, and the steady state is 31.4A and 15.8A; the average current overshoot before feedforward compensation is 7.9% and 13.4%, the adjustment time is 2.4e-6 and 2.4e-6, and the steady state is 40.0A and 20.3A. Under the constant current step of 50%-75%-50%, the average current overshoot before feedforward compensation is 13.3% and 22.8%, the adjustment time is 3.6e-6 and 4.0e-6, and the steady state is 47.1A and 31.6A; the average current overshoot before feedforward compensation is 5.0% and 7.6%, the adjustment time is 1.6e-6 and 2.4e-6, and the steady state is 60.0A and 40.3A. Under the rated current step of 10%-90%-10%, the average current overshoot before feedforward compensation is 32.2% and 50.2%, the adjustment time is 4.4e-6 and 1.97e-4, and the steady state is 56.2A and 7.48A; the average current overshoot before feedforward compensation is 8.4% and 37.3%, the adjustment time is 2.4e-6 and 1.99e-4, and the steady state is 71.8A and 9.18A.

[0123] It can be seen from multiple sets of data that after adding feedforward control, the digital control system has better steady-state tracking performance and dynamic performance than before the compensation was added, especially the dynamic performance improvement under large load steps.

[0124] The delay link and sample-and-hold circuit of a digital system are usually equivalent to one beat, so the simulation mainly verifies the sensitivity of the improved deadbeat control algorithm to the DC inductance parameters. L Take 0.5, 0.75, 1 and 2 for simulation verification respectively. Figure 7 The step response of the average inductor current of the phase-shifted full-bridge converter without feedforward compensation is shown at different model matching degrees.

[0125] Depend on Figure 7 It can be seen that for a step input signal, K L When it is equal to 0.5, the inner current loop is in a critical oscillation state and cannot converge. <K L <1, the system converges, and the average current has overshoot during the transition process. L The increase of K reduces the system overshoot. L ≥1, the system has no overshoot, as K L As the current increases, the response speed of the inner loop slows down.

[0126] Considering that the duty cycle loss of the phase-shifted full-bridge converter will affect the performance of the system, the duty cycle of the system is compensated and the robustness of the compensated system is analyzed.

[0127] Depend on Figure 8It can be seen that after duty cycle feedforward compensation, the system can accurately track the current reference value, there is no steady-state error, and the system dynamic performance is better than the uncompensated system. In particular, when the actual inductance value is equal to the model inductance value (K L =1), the system response speed is optimal. Compared with the dynamic effect before compensation, Figure 8 In (a), the system after duty cycle compensation is K L =0.5, it can also converge and stabilize. Through simulation, it can be obtained that the model matching parameter K under this circuit parameter L The stability margin can be as low as 0.41, so the robustness of the system can be further improved through duty cycle feedforward compensation.

[0128] In order to verify the feasibility and effectiveness of the predictive zero-beat average model control method, a phase-shifted full-bridge converter test device was built. Except for a 3μf film capacitor connected in series at the input end of the high-frequency transformer to prevent the transformer from being magnetized, the other circuit parameters were consistent with the simulation parameters.

[0129] The control system of the phase-shifted full-bridge converter uses STM32F103 as the core, and the peripheral expansion power supply circuit, analog sampling circuit, drive circuit and other circuits. The control method is implemented by software. The inductor current oversampling technology uses direct memory access (DMA) and the built-in analog-to-digital converter (ADC) of STM32F103 to achieve collaborative implementation. DMA transmission can copy data directly from one memory space to another without the CPU participating in the control, which greatly improves the CPU utilization efficiency. Every time the sampling is completed, DMA automatically stores the sampling results into an array unit, and the number of the array unit increases in sequence each time the sampling is completed. Thanks to this feature of DMA, the ADC can work at a very high sampling rate. The built-in ADC of STM32F103 supports sampling once every 1.55us at most. The phase-shifted full-bridge converter running at a switching frequency of 25kHz can sample 28 times in one switching cycle, which is much higher than the switching frequency and is very suitable for the deadbeat average model control.

[0130] Fig. 9 and Fig.10The inductor current waveforms of the traditional deadbeat control and the predicted deadbeat average model control methods under different current settings are shown respectively. During the test, the program removes the voltage outer loop, and the current loop directly uses a constant digital setting. A resistor of about 0.7 ohms is connected to the power output. From the figure, it can be seen that the peak-to-peak value of the predicted deadbeat average model control method is significantly smaller than the peak-to-peak value of the traditional deadbeat control method. The peak-to-peak value of the predicted deadbeat average model control method is 6.3A and 9.6A under the current settings of 30A and 60A, while the peak-to-peak value of the traditional deadbeat ripple is 10A and 15A. And from the detailed waveform in the lower right corner of the figure, it can be observed that the peak value and valley value of the predicted deadbeat average model control method are almost stably working on a straight line.

[0131] At the same time, during the test, the oscilloscope level trigger function captured the occasional impact fluctuation of the inductor current during the traditional deadbeat control. However, this phenomenon was not found when the predictive deadbeat average model control method was used, which also proves that the predictive deadbeat average model control method has stronger anti-interference and stability.

[0132] Fig.11 and Fig.12 The inductor current waveforms of the traditional deadbeat control and the predicted deadbeat average model control method are shown respectively under the current loop step setting of 30A-60A-30A. During the experiment, the voltage outer loop is removed, and the current loop setting directly adopts the step setting value. The step waveform of channel 4 in the figure represents the time point of the current step command change, and a resistor of about 0.7 ohms is connected to the output end of the power supply. It can be found from the figure that the traditional deadbeat has a large overshoot in the dynamic process because it does not adopt predictive control, and it takes multiple oscillations to stabilize. The predicted deadbeat average model control method has good dynamic performance, and it almost tracks the given signal within one control cycle and has basically no overshoot. At the same time, in the process of the output voltage gradually rising or falling, the predicted deadbeat average model control has a good tracking effect on the inductor current and is better than the traditional deadbeat control.

[0133] Fig.13 and Fig.14 The inductor current waveform and output voltage waveform of the traditional deadbeat control and the predicted deadbeat average model control method under different loads are shown when the output voltage is set to 50V by connecting to the voltage outer loop. It can be seen from the figure that the peak-to-peak value of the current of the predicted deadbeat average model control method is significantly smaller than the peak-to-peak value of the current of the traditional deadbeat control method. The peak-to-peak value of the current ripple of the predicted deadbeat average model control method is 8A and 12A under 32A and 78A loads, but the peak-to-peak value of the current ripple of the traditional deadbeat control is 14A and 17A. The experimental phenomenon verifies that the predicted deadbeat average model control method matches the voltage outer loop control well and meets the system control voltage regulation performance requirements.

[0134] According to the embodiments of the present invention as described above, these embodiments do not describe all the details in detail, nor do they limit the invention to the specific embodiments described. According to the above description, many modifications and changes can be made. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can make good use of the present invention and the modified use based on the present invention. The present invention is only limited by the claims and their full scope and equivalents.

Claims

1. A current prediction deadbeat average model control method for a phase-shifted full-bridge converter, characterized in that: The control method is used on a phase-shifted full-bridge converter, which is sequentially connected by a full-bridge inverter circuit, a resonant inductor, a high-frequency power transformer, a full-bridge rectifier circuit, and an LC output filter circuit. It adopts a double closed-loop control system, in which the voltage outer loop adopts PI control and the current inner loop adopts a predicted deadbeat average model control method. The predicted deadbeat average model control method is: When the system runs in cycle k, the output filter inductor current is oversampled and averaged to obtain the average value of the inductor current in cycle k, I avg (k); When the system runs in the k+1 cycle, according to the average value of the inductor current I avg (k) and k+2 cycle current given value I set (k+2) calculates the control signal D of k+2 period k+2 , the calculation formula is: Among them, T s is the control period, L is the output filter inductance, D k is the k-cycle duty cycle, D k+1 is the duty cycle of the k+1 cycle obtained by the above formula in the k-1 cycle, K is the primary-secondary ratio of the high-frequency power transformer, U in is the input voltage of the phase-shifted full-bridge converter, U out is the output voltage of the phase-shifted full-bridge converter, I set (k+1) is the given current value of the k+1 cycle.

2. The control method according to claim 1, characterized in that: The control method adopts feedforward compensation to compensate the duty cycle to overcome the influence of the duty cycle loss of the secondary side of the transformer of the phase-shifted full-bridge converter on the control effect.

3. The control method according to claim 2, characterized in that: In the feedforward compensation, the secondary side lost duty cycle signal that needs to be compensated for k cycles can be expressed as: Among them, L r It is the sum of the resonant inductance and the leakage inductance of the high-frequency power transformer T.

Images