A sliding mode predictive control method for wide output DC-DC converter

Abstract

The application belongs to the technical field of converter control, and provides a sliding mode prediction control method of a wide output DC-DC converter, comprising: collecting input voltage, output voltage and inductor current of the converter in real time; obtaining an inductor current reference value under ideal steady state according to a preset target reference voltage value; obtaining a discrete state prediction model at a next control time, including a predicted inductor current and a predicted output voltage; introducing an adaptive sliding film surface function, and deriving a sliding film surface state prediction value at the next control time; constructing a cost function; traversing switch combinations, selecting a switch state sequence corresponding to a minimum value of the cost function, and converting into a PWM duty cycle signal for control. The application fuses the strong robustness of the sliding mode control and the multi-target rolling optimization mechanism of the model prediction control, realizes smooth output in a wide voltage range, completely eliminates sliding mode chattering, and significantly improves the ultrafast dynamic response ability of the system to sudden loads.

Description

Technical Field

[0001] This invention relates to the field of converter control technology, and more specifically to a sliding mode predictive control method for a wide-output DC-DC converter. Background Technology

[0002] With the rapid development of a new generation of intelligent equipment, represented by humanoid robots and high-dynamic industrial servo systems, their internal power supply systems place stringent demands on DC-DC converters, requiring both "wide voltage regulation" and "extremely high dynamic response." For example, when humanoid robots perform high-burst actions such as jumping and running, not only is it required that the DC-DC converter be able to smoothly switch between boost and buck modes over a wide input / output voltage range, but their servo motors also generate extremely drastic load current surges.

[0003] Wide-output DC-DC converters (such as four-switch Buck-Boost converters) often employ traditional dual-closed-loop PI control. However, PI control is designed based on a small-signal model of local operating point. When faced with wide-range voltage regulation and large nonlinear disturbances, parameter mismatch often occurs, leading to excessive system voltage overshoot or even instability. The inherent linear assumption of its transfer function determines that the PI controller cannot cope with the global nonlinear dynamics across modes (Buck / Boost switching).

[0004] To address nonlinearity and disturbance rejection issues, sliding mode control (SMC) has been widely adopted. SMC exhibits strong global robustness to changes in internal system parameters and external load disturbances; however, its inherent discontinuous switching characteristics can lead to severe high-frequency chattering. Specifically, the control law of a traditional first-order sliding mode controller typically includes the sign function sgns(u=ueq+K). sgns), where ueq is the equivalent control quantity, K is the switching gain, and s is the sliding surface variable. Due to the discontinuity of sgns, once the system state reaches the sliding surface, the control quantity varies between +K and... Switching between K modes at a theoretically infinite frequency leads to: a sharp increase in switching losses of the power switching transistor; severe electromagnetic interference (EMI) and inductor howling; and severely limits its application in practical high-frequency, high-power-density power supply systems. Although it can reduce chattering to some extent, the design complexity of its high-order sliding surface is high, and the convergence speed and accuracy of the observer are limited by the uncertainty range of the system. Therefore, its engineering practicality in wide-output multi-mode switching scenarios still has room for improvement.

[0005] Finite Control Set Model Predictive Control (FCS-MPC), as an advanced discrete-time domain optimization strategy, has attracted much attention in recent years. MPC elegantly handles multivariable cooperative constraints through online rolling optimization and achieves extremely fast transient response without the need for traditional PWM modulators. However, existing MPC techniques have a fatal flaw: their prediction accuracy is highly dependent on the precise discrete mathematical model of the controlled object. In wide-output applications, physical parameters such as inductance L and capacitance C drift significantly with changes in bias current (core saturation effect), and rapid load steps introduce strong unmodeled disturbances. Once parameter mismatch occurs, the prediction results of traditional MPC will deviate significantly from the actual state, leading to increased steady-state error, current distortion, and even control failure.

[0006] In summary, existing technologies exhibit both challenges: sliding mode control is "disturbance resistant but suffers from severe chattering," and model predictive control is "extremely fast but highly dependent on accurate models." Neither can perfectly meet the stringent requirements of wide-output DC-DC converters under extremely complex operating conditions. This invention aims to overcome the limitations of traditional single control algorithms by deeply integrating the global robustness of sliding mode control with the rolling optimization mechanism of model predictive control, thereby simultaneously addressing the two core pain points mentioned above. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a sliding mode predictive control method for a wide-output DC-DC converter, thereby overcoming the limitations of a single control algorithm and improving the system's rapid dynamic response capability in the face of sudden load changes.

[0008] This invention provides a sliding mode predictive control method for a wide-output DC-DC converter, comprising: The input voltage, output voltage, and inductor current of the converter are acquired in real time; and the inductor current reference value under ideal steady state is obtained based on the preset target reference voltage value. Obtain the discrete state prediction model for the next control moment, including predicting the inductor current and the output voltage; An adaptive sliding surface function is introduced, and the predicted value of the sliding surface state at the next control moment is obtained by deduction. Construct the cost function; the cost function includes the sliding mode convergence term, the switching frequency penalty term, and the inductor current hard constraint penalty function. The switch combinations are iterated, the switch state sequence corresponding to the minimum value of the cost function is selected, and then converted into a PWM duty cycle signal for control.

[0009] Optionally, under ideal steady-state conditions, when the converter output voltage reaches the target reference voltage value... At that time, i.e., output voltage And taking into account the switching function of the input-side half-bridge under ideal steady state. duty cycle Then the inductor current reference value R is the system equivalent load resistance. This is the input voltage.

[0010] Optionally, the discrete state prediction model is: ,in To predict inductor current, To predict the output voltage, L is the value of the energy storage inductance, and C is the output filter capacitor. The set system sampling period; and These are the switching functions for the input-side half-bridge and the output-side half-bridge, respectively.

[0011] Optionally, adaptive sliding surface function ,in, and These are the sliding surface weighting coefficients for voltage error and current error, respectively. Voltage error Current error ; Then the predicted value of the sluice surface state at the next control moment .

[0012] The introduction of discretized sliding surface prediction state: The constructed "discrete sliding surface prediction state" s(k+1) is directly used as the tracking target of the model prediction control cost function, replacing the tracking method based on absolute error in traditional MPC.

[0013] Optionally, based on the switching state of the FSSB topology, the predicted values ​​of the sliding surface state at the next control moment are as follows: when , ; when , ; when , ; when , .

[0014] Optionally, the cost function is , For sliding mode approaching terms; For the switching frequency penalty term, where These are constant weighting coefficients; This is the hard constraint penalty function for inductor current.

[0015] In summary, sliding mode robustness (first term), switching frequency optimization (second term), and hardware safety constraints (third term) are integrated into a single cost function, achieving multi-objective collaborative optimization.

[0016] By adopting the above technical solution, this application has the following beneficial effects: This invention breaks through the limitations of traditional single control algorithms, deeply integrating the strong robustness of sliding mode surfaces with the multi-objective rolling optimization mechanism of model predictive control (MPC), resulting in extremely significant beneficial technical effects. First, compared to traditional sliding mode control (SMC), this invention abandons the hard-switching logic of the sign function that causes high-frequency chattering, using the discretized sliding mode surface prediction value as the optimization tracking target, and introducing a switching action penalty term in the composite cost function, fundamentally achieving "chatter-free" smooth control, which can significantly reduce the switching losses and electromagnetic interference of power devices. Second, compared to traditional model predictive control (MPC), this invention completely overcomes its performance bottleneck of being extremely dependent on accurate mathematical models. Even when facing extreme load steps such as motor emergency stop and start-up, or encountering physical parameter drifts such as ±30% inductance, the natural invariance of the sliding mode surface still ensures extremely small transient voltage overshoot and microsecond-level extremely fast dynamic recovery speed, greatly improving the system's global disturbance rejection capability. Finally, compared with traditional dual-loop PI control, this invention perfectly adapts to wide-output multi-switch topologies, eliminating the need for cumbersome cascaded loops and segmented parameter tuning. It achieves smooth and seamless switching between buck and boost modes across the entire voltage range using a single cost function. In summary, this invention achieves comprehensive breakthroughs in transient response, resistance to parameter mismatch, and smoothness of wide-range voltage regulation. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0018] Figure 1 A schematic diagram of the FSBB topology provided in an embodiment of the present invention is shown; Figure 2 One of the flowcharts of a sliding mode predictive control method for a wide-output DC-DC converter provided by an embodiment of the present invention is shown; Figure 3 This is a second flowchart illustrating a sliding mode predictive control method for a wide-output DC-DC converter provided by an embodiment of the present invention; Figure 4 A comparison diagram of the output voltage and inductor current of the method provided in the embodiments of the present invention with those of conventional PI, SMC and MPC is shown; Figure 5 A schematic diagram showing details of the transient response to a sudden load change (180Ω→90Ω) provided in an embodiment of the present invention is shown; Figure 6 A schematic diagram of the trajectory of the sliding surface variable s(k) provided in an embodiment of the present invention is shown; Figure 7 A schematic diagram showing the comparison of steady-state voltage ripple provided in an embodiment of the present invention is shown; Figure 8 A schematic diagram illustrating the robustness verification of ±30% inductance parameter perturbation provided in an embodiment of the present invention is shown. Figure 9 This diagram illustrates the seamless switching between Buck and Boost modes provided by an embodiment of the present invention. Detailed Implementation

[0019] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are only used to more clearly illustrate the technical solution of the present invention and are therefore merely examples, and should not be construed as limiting the scope of protection of the present invention. It should be noted that, unless otherwise stated, the technical or scientific terms used in this application should have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0020] To address the issues of poor parameter mismatch resistance in conventional model predictive control (MPC) and high-frequency chattering in conventional sliding mode control (SMC) when facing drastic load changes in existing wide-output DC-DC converters, a sliding mode predictive control method for wide-output DC-DC converters is provided in one embodiment. The implementation of the method provided in this embodiment is illustrated below using a four-switch Buck-Boost (FSBB) topology as an example.

[0021] First, the four-switch buck-boost (FSBB) topology will be explained. Figure 1 As shown, this topology consists of an input-side half-bridge ( , ) and output half-bridge ( , It consists of two half-bridges connected by an energy storage inductor L. A filter capacitor C and a load resistor are connected in parallel on the output side. Where: - and Construct high-side and low-side switches on the input side, and define the switching function. (1 indicates) The upper tube is conducting, 0 indicates (Lower pipe conduction); - and Construct the high-side and low-side switches on the output side, and define the switching functions. ∈{0,1} (1 represents The upper tube is conducting, 0 indicates (Lower transistor conducts). The FSBB converter passes through... and Different combinations achieve three working modes, as shown in Table 1: Table 1

[0022] when When the converter operates in buck mode; when At that time, it operates in Boost mode; when At that time, it operates in a hybrid boost / buck mode.

[0023] Based on this, such as Figure 2-3 As shown, the sliding mode predictive control method for a wide-output DC-DC converter provided in this embodiment includes: S1. Real-time acquisition of the converter's input voltage, output voltage, and inductor current; and obtaining the inductor current reference value under ideal steady state based on the preset target reference voltage value.

[0024] To fully capture the real-time operating characteristics of the wide-output four-switch Buck-Boost (FSBB) converter, in the current k-th control cycle (with the system sampling period set to 1), The input voltage of the converter is acquired in real time through a high-precision analog-to-digital converter (ADC). Output voltage and inductor current Simultaneously, the target reference voltage set by the host computer or the outer loop is acquired. In variable structure control, to ensure the power balance and global stability of the system, the steady-state energy conservation law under ideal conditions (i.e., input power equals output power) is applied. ,Right now Ignoring internal parasitic losses, a reference current calculation mechanism is introduced. For the FSBB converter, under steady-state CCM (continuous current mode), the average inductor current... The relationship with the input current depends on the duty cycle. ( Duty cycle): .

[0025] Under ideal steady-state conditions, when the converter output voltage reaches the reference value... At that time, Substituting into the energy conservation equation, and considering the steady state... (Using Buck mode approximation), the reference value of the inductor current matching the target voltage is calculated. The calculation formula (1) is as follows: (1) In the formula, R is the equivalent load resistance of the system. Through the solution of formula (1), the voltage is established at the beginning of a single cycle. With current The dual reference datum provides extremely precise setting input values ​​for the multi-objective sliding surface design in subsequent steps.

[0026] S2. Obtain the discrete state prediction model for the next control moment, including predicting the inductor current and the output voltage.

[0027] The original single-variable linear control is insufficient to handle the nonlinear dynamic behavior of wide-output converters when switching between boost and buck modes. First, a continuous-time state-space model is established based on Kirchhoff's voltage law (KVL). The loop equation (2) for inductor L is: (2) when ( (Conduction), input voltage Applied to the left end of the inductor; when ( (Conducting), the left end of the inductor is grounded; when ( (Conductivity), the right end of the inductor is connected to the output voltage. ;when ( (Conductivity), the right end of the inductor is grounded; therefore, the voltage formula (3) across the inductor is: (3) According to Kirchhoff's Current Law (KCL), the current equation (4) for the output node is as follows: (4) Inflow to output capacitor and load The current in the parallel branch comes from the inductor, but only when ( When the circuit is turned on, the inductor current flows to the output side; the load current is... Therefore, the capacitor current Define the state vector. Control input The above equation can be expressed in standard state-space form (5):

[0028] in:

[0029] To accommodate the discrete execution characteristics of digital microcontrollers (DSPs), the first-order forward Euler method is used to perform differential discretization on the differential terms. For general differential equations... The forward Euler discretization is as follows:

[0030] Right now:

[0031] The stability requirement of the forward Euler method is the sampling period. Small enough, that is To ensure that the eigenvalues ​​of the discrete system fall within the unit circle. Based on the sampled values ​​at the current time k, the discrete state prediction model of the system at the next control time k+1 is derived. Substituting equations (6) and (7) into the forward Euler formula, the inductor current prediction formula (8) and the output voltage prediction formula (9) are obtained as follows: (8) (9) In the formula, L is the value of the energy storage inductance; C is the value of the output filter capacitor. Through formulas (8) and (9), the model can predict in advance the evolution trajectory of inductor current and output voltage under different switching combinations, providing basic feature deduction capabilities for subsequent model prediction.

[0032] S3. Introduce an adaptive sliding surface function and deduce the predicted value of the sliding surface state at the next control moment.

[0033] Traditional PID controllers are prone to failure when faced with large nonlinear disturbances such as the rapid start and stop of motors in humanoid robots. To balance the precise voltage regulation of the converter over a wide output range with the rapid current limiting under sudden load changes, this invention introduces the concept of linear sliding mode variable structure control.

[0034] Based on the reference obtained in step S1, the output voltage tracking error and inductor current tracking error of the system at time k are defined respectively, as follows:

[0035]

[0036] Therefore, an adaptive sliding surface function s(k) containing dual error characteristics is constructed, and its formula (12) is as follows:

[0037] In the formula, and The sliding surface weights for voltage error and current error are respectively, and their values ​​must be greater than zero. They are used to determine the relative convergence speed of the system state on the phase trajectory. Through the sliding surface constructed by formula (12), the system state is forcibly bound to the preset manifold. By utilizing the inherent invariance of sliding mode control, it effectively resists the interference caused by external load jumps and internal component parameter drifts, greatly enhancing robustness.

[0038] S4. Construct the cost function; the cost function includes the sliding mode convergence term, the switching frequency penalty term, and the inductor current hard constraint penalty function.

[0039] Traditional sliding mode control relies solely on the error at the current moment. Triggering control actions constitutes a delayed intervention. The core breakthrough of this invention lies in the mathematical fusion of the predictive model and sliding mode control on a time scale.

[0040] Pushing the time scale of the sliding surface function forward to step k+1, its formula (13) is:

[0041] The discrete prediction feature variables extracted in step S2 (Formula (8)) and (Formula (9)) is directly substituted into the derived sliding surface function. After mathematical expansion and simplification, the predicted value s(k+1) of the sliding surface state in the next cycle under different switching state combinations is derived, and its formula (14) is: (14) Formula (14) reveals the predicted value of the sliding surface. With current sliding surface value The recursive relationship between them clarifies the different switch combinations Quantitative impact on the future evolution trajectory of the sliding surface. This is the mathematical hub connecting model prediction and variable structure control, enabling the controller to accurately calculate the quantitative impact of different current switching actions on the future sliding trajectory of the system, providing a data foundation for the subsequent complete replacement of traditional sliding mode hard switching logic.

[0042] For the four effective switching combinations of the FSBB converter The predicted values ​​are as follows: (1) : (15a) (2) : (15b) (3) : (15c) (4) : (15d) S5. Traverse the switch combinations, select the switch state sequence corresponding to the minimum value of the cost function, and convert it into a PWM duty cycle signal for control.

[0043] Traditional sliding mode control relies on a sign function to force state switching. This hard-switching logic... The nearby area causes high-frequency oscillations in the control quantity, a phenomenon known as "beatle".

[0044]

[0045] To overcome the pain point of chattering, this invention reconstructs the algorithm evaluation mechanism and constructs a composite cost function J with the future sliding surface prediction value as the core tracking target.

[0046] To force the system state trajectory to approach the sliding mode origin at the fastest speed in the next cycle (i.e. At the same time, it actively suppresses high-frequency jitter and protects hardware overload. The cost function is composed of three penalty terms that are linearly superimposed. The following composite cost calculation formula (16) is designed: (16) In the formula, the cost function is composed of a linear superposition of three penalty terms: First item For the sliding mode approaching term, the calculation result of formula (14) is used. Take the absolute value to ensure the strong robustness of the closed-loop system. Minimize This is equivalent to forcing the system state to approach the sliding surface origin s=0 as quickly as possible in the next cycle. Unlike the hard switching of traditional sliding mode, this approach naturally selects the optimal switching state through "soft optimization" of the cost function, avoiding... discontinuity; The second term is the switching frequency penalty term, which is determined by introducing a constant weighting coefficient. >0 and combine it with the historical switch state from the previous moment. (j=1, 2 correspond to the input and output half-bridges respectively), an additional cost penalty is imposed on the switching sequence that produces meaningless state transitions. When When, the contribution is 0; when At that time, the contribution was . The selection requires a trade-off between dynamic response speed and switching losses: Excessive: The system tends to maintain the current on / off state, resulting in slower dynamic response; Too small: The switching frequency penalty is insufficient and cannot effectively suppress meaningless switching; Third item This is a hard constraint penalty function for inductor current. When the predicted current exceeds the threshold, the cost function of the corresponding switch combination tends to infinity, thus being automatically excluded during the optimization process, thereby achieving hardware overcurrent protection.

[0047] In the final stage of the algorithm execution, the controller transforms the optimization process into a minimum search problem on a discrete finite set. For the FSBB converter, its valid physical control set U contains only 4 sets of valid states: .

[0048] The digital controller iterates through the four switch combinations online within a very short control cycle, calculating the corresponding cost function J for each combination. The adaptive selection of the optimal switch state sequence follows the principle of minimization, expressed by formula (17): (17) in, These are candidate switching states; the superscript opt ​​indicates the optimal solution after optimization calculation, that is, the optimal combination of switching state components that makes the cost function J reach the global minimum within the current control cycle. and These represent the optimal control commands for the output and input half-bridges at the next time step, respectively. arg min represents the value of the independent variable that minimizes the cost function.

[0049] The controller selects the unique combination of switching components that minimizes the global cost function J and converts it into a pulse width modulation (PWM) duty cycle signal, which is then sent to the underlying driver circuit. Subsequently, the system updates the historical switching states and discrete variables, sets k=k+1, and enters the online rolling optimization for the next control cycle. Through this extremely agile discrete space optimization mechanism, this invention exhibits significantly greater flexibility and system stability than traditional linear algorithms in a wide range of continuous voltage regulation scenarios.

[0050] Based on the above scheme, this embodiment breaks through the limitations of traditional single control algorithms, deeply integrating the strong robustness of the sliding mode surface with the multi-objective rolling optimization mechanism of model predictive control (MPC), resulting in extremely significant beneficial technical effects. First, compared with traditional sliding mode control (SMC), this invention abandons the hard-switching logic of the sign function that causes high-frequency chattering, uses the discretized sliding mode surface prediction value as the optimization tracking target, and introduces a switching action penalty term in the composite cost function, fundamentally achieving "chatter-free" smooth control, which can significantly reduce the switching losses and electromagnetic interference of power devices. Second, compared with traditional model predictive control (MPC), this invention completely breaks through its performance bottleneck of being extremely dependent on accurate mathematical models. Even when facing extreme load steps such as motor emergency stop and start-up, or encountering ±30% drift in physical parameters such as inductance, the natural invariance of the sliding mode surface can still guarantee extremely small transient voltage overshoot and microsecond-level extremely fast dynamic recovery speed, greatly improving the system's global disturbance rejection capability. Finally, compared with traditional dual-loop PI control, this invention perfectly adapts to wide-output multi-switch topologies, eliminating the need for cumbersome cascaded loops and segmented parameter tuning. It achieves smooth and seamless switching between buck and boost modes across the entire voltage range using a single cost function. In summary, this invention achieves comprehensive breakthroughs in transient response, resistance to parameter mismatch, and smoothness of wide-range voltage regulation.

[0051] To fully verify the technical effectiveness of the sliding mode predictive control method proposed in this invention, the following is combined with... Figures 4 to 9 The simulation and experimental test results of the embodiments of the present invention are described in detail. The simulation was conducted using an FSBB topology model built in the MATLAB / Simulink environment, with the main parameters set as follows: input voltage V... in =24V, target output voltage V ref Adjustable from 12V to 36V, energy storage inductor L=200μH, output filter capacitor C=470μF, rated load resistance R=180Ω, system sampling period Ts=10μs, sliding surface weighting coefficients c1=1, c2=0.05, switching frequency penalty weighting. λ sw =0.02. The method proposed in this invention was compared with traditional PI, SMC and MPC methods under the same operating conditions.

[0052] Figure 4 A comparison graph showing the output voltage and inductor current of the method provided in this embodiment of the invention with those of conventional PI, SMC, and MPC methods is presented. Figure 4 It can be seen that during the system startup to steady-state phase, the output voltage V under traditional PI control... out There is an overshoot of approximately 18%, and the settling time is as long as about 5ms; although traditional SMCs have a relatively fast response speed, the steady-state inductor current I... LThe method exhibits significant high-frequency chattering, with peak-to-peak ripple reaching ±0.5A. While traditional MPC performs well under nominal parameters, its transient waveform recovery is still inferior to the method proposed in this invention. In contrast, the proposed method controls the output voltage overshoot to within 2%, has a settling time of less than 1.5ms, and produces a smooth inductor current trajectory without significant chattering, fully demonstrating the comprehensive advantages of the deep integration of sliding mode robustness and model prediction rolling optimization.

[0053] Figure 5 A schematic diagram illustrating the transient response details of a sudden load change (180Ω→90Ω) provided in an embodiment of the present invention is shown. At t=20ms, the load resistance jumps from 180Ω to 90Ω, equivalent to an instantaneous doubling of the output current, simulating extreme conditions such as a sudden start-up of a servo motor. Figure 5 As can be seen, traditional PI control experiences a momentary voltage drop of approximately 1.6V and a recovery time exceeding 3ms due to the mismatch in the small-signal linearization model. While traditional SMC exhibits good robustness, it re-enters high-frequency chattering in steady state. Traditional MPC experiences a voltage drop of approximately 0.8V and a recovery time of approximately 1.2ms at the nominal parameter level. The method proposed in this invention achieves a voltage drop of only approximately 0.3V and a recovery time of less than 200μs, reaching microsecond-level dynamic recovery. This fully demonstrates the extremely fast dynamic response capability brought about by using the discretized sliding surface prediction value as the optimization tracking target.

[0054] Figure 6 A schematic diagram of the trajectory of the sliding surface variable s(k) provided in an embodiment of the present invention is shown. Figure 6 As can be seen, under the traditional SMC method, the sliding surface variable s(k) exhibits high-frequency jumps near its zero value, with a jump amplitude of approximately ±0.4, corresponding to frequent hard switching of the control quantity between ±K, which is the root cause of chattering. However, under the method proposed in this invention, the sliding surface variable s(k) smoothly converges from its initial value to an extremely narrow band (bandwidth of approximately ±0.05) near its zero value, without any high-frequency jumps. This is because this invention replaces the hard switching logic of the traditional sgn(s) with a "soft optimization" mechanism of the cost function J, and actively suppresses meaningless state flips through a switching frequency penalty term, thereby fundamentally eliminating sliding mode chattering.

[0055] Figure 7 A schematic diagram illustrating the steady-state voltage ripple comparison provided by an embodiment of the present invention is shown. Figure 7 As can be seen, the peak-to-peak value of the output voltage ripple ΔV under the traditional SMC method ppThe ripple of the traditional MPC method is approximately 120mV, with a cluttered high-frequency spectrum and severe EMI noise; the ripple of the traditional method is approximately 80mV; while the steady-state voltage ripple peak-to-peak value of the method proposed in this invention is reduced to less than 30mV, with a concentrated spectrum and clean waveform. This is because the switching frequency penalty term in the cost function, while ensuring dynamic response, constrains the average switching frequency within a reasonable range (approximately 30kHz~50kHz), effectively reducing the switching losses and electromagnetic interference of the power switching devices.

[0056] Figure 8 This diagram illustrates the robustness verification of ±30% inductance parameter perturbation provided in an embodiment of the present invention. Considering the core saturation effect, the actual value of the energy storage inductor L is perturbed to +30% (260μH) and -30% (140μH) of the nominal value, respectively, and the steady-state and dynamic performance of each method is observed. Figure 8 It is evident that traditional MPC methods, due to their heavy reliance on accurate discrete models, exhibit significant steady-state current deviations and increased voltage ripple when parameters are mismatched, with the maximum steady-state deviation exceeding 5%. In contrast, the method proposed in this invention, under L-parameter ±30% perturbation conditions, consistently maintains an output voltage steady-state deviation of less than 1%, and the transient overshoot and recovery time remain almost identical to those under nominal parameters. This fully demonstrates that, by leveraging the predictive tracking mechanism of the "natural invariance" of the sliding mode surface, this invention completely overcomes the performance bottleneck of traditional MPC's "heavy reliance on accurate models," significantly improving global resistance to parameter mismatch.

[0057] Figure 9 This diagram illustrates the seamless Buck→Boost mode switching provided by an embodiment of the present invention. At t=40ms, the target reference voltage V... ref The voltage jumps from 18V (Buck mode) to 30V (Boost mode), crossing the boundary between boost and buck modes. Figure 9 As can be seen, traditional PI control, due to limitations imposed by cascaded loops and segmented parameter tuning, exhibits significant voltage dips and current surges during mode switching, with a transition time of approximately 2ms. In contrast, the method proposed in this invention utilizes an online optimization mechanism for the entire switching combination driven by a single cost function to achieve smooth and seamless switching between Buck and Boost modes. The voltage transition curve is continuous and monotonic, with no visible dips or surges, and a transition time of less than 500μs. This demonstrates that this invention perfectly adapts to wide-output multi-switch topologies, eliminating the need for cumbersome cascaded loops and segmented parameter tuning, and enabling smooth and seamless voltage regulation across the entire voltage range.

[0058] comprehensive Figures 4 to 9 The simulation test results show that the sliding mode predictive control method for wide-output DC-DC converters proposed in this invention is significantly better than traditional PI, SMC, and MPC methods in many key indicators such as transient response speed, steady-state accuracy, resistance to parameter mismatch, suppression of chattering, and smoothness of wide-range voltage regulation, demonstrating outstanding technical effects.

[0059] The above embodiments are only used to provide a detailed description of the technical solutions of this application. However, the descriptions of the above embodiments are only for the purpose of helping to understand the methods of the embodiments of the present invention and should not be construed as limiting the embodiments of the present invention. Any variations or substitutions that can be easily conceived by those skilled in the art should be covered within the protection scope of the embodiments of the present invention.

Claims

1. A sliding mode predictive control method for a wide-output DC-DC converter, characterized in that, The wide-output DC-DC converter is an FSBB topology, including an input-side half-bridge and an output-side half-bridge, which are connected via an energy storage inductor. The output-side half-bridge has a filter capacitor and a load resistor connected in parallel. The method includes: The input voltage, output voltage, and inductor current of the converter are acquired in real time; and the inductor current reference value under ideal steady state is obtained based on the preset target reference voltage value. Obtain the discrete state prediction model for the next control moment, including predicting the inductor current and the output voltage; An adaptive sliding surface function is introduced, and the predicted value of the sliding surface state at the next control moment is obtained by deduction. Construct the cost function; the cost function includes the sliding mode convergence term, the switching frequency penalty term, and the inductor current hard constraint penalty function. The switch combinations are iterated, the switch state sequence corresponding to the minimum value of the cost function is selected, and then converted into a PWM duty cycle signal for control.

2. The method according to claim 1, characterized in that, Under ideal steady-state conditions, when the converter output voltage reaches the target reference voltage value... At that time, i.e., output voltage And taking into account the switching function of the input-side half-bridge under ideal steady state. duty cycle Then the inductor current reference value R is the system equivalent load resistance. This is the input voltage.

3. The method according to claim 2, characterized in that, Discrete state prediction model is ,in To predict inductor current, To predict the output voltage, L is the value of the energy storage inductance, and C is the output filter capacitor. The set system sampling period; and These are the switching functions for the input-side half-bridge and the output-side half-bridge, respectively.

4. The method according to claim 3, characterized in that, Adaptive sliding surface function ,in, and These are the sliding surface weighting coefficients for voltage error and current error, respectively. Voltage error Current error ; Then the predicted value of the sluice surface state at the next control moment .

5. The method according to claim 4, characterized in that, Based on the switching state of the FSSB topology, the predicted values ​​of the sliding surface state at the next control moment are as follows: when , ; when , ; when , ; when , .

6. The method according to claim 4, characterized in that, The cost function is , For sliding mode approaching terms; For the switching frequency penalty term, where The weighting coefficients are constants. This is the hard constraint penalty function for inductor current.

Images