Superspiral sliding mode active disturbance rejection control strategy based on four active bridge converters

Abstract

The invention relates to a superspiral sliding mode active disturbance rejection control strategy based on a four-active bridge converter, and the method comprises the following steps: employing an extended state observer to estimate uncertain factors, unmodeled dynamics and coupling terms in a system, putting the disturbance estimation into superspiral sliding mode closed-loop control for elimination, and independent control of voltages of all ports is realized. After voltage and phase shift of each port of the four-active bridge converter are input into an extended state observer to estimate total disturbance of a system, given voltage is input into a differential tracker to eliminate influence of sudden change of the given voltage on a control effect, and obtained tracking voltage and observed voltage of the observer are input into a subtractor for subtraction; the method comprises the following steps of: inputting the obtained error quantity into a sliding mode controller, obtaining a phase shift ratio di through corresponding operation, giving phase shift ratios d2, d3 and d4 between primary and secondary side ports, generating corresponding PWM (Pulse-Width Modulation) signals, driving corresponding switching tubes to be switched on and switched off, and realizing independent voltage control and more efficient and reliable control operation of each port.

Description

Technical Field

[0001] This invention belongs to the field of power electronics technology, specifically relating to a super-spiral sliding mode active disturbance rejection control strategy based on a four-active-bridge converter. Background Technology

[0002] Dual-Active Bridge (DAB) converters have found practical applications in renewable energy and electric vehicle charging stations, offering current isolation and zero-voltage switching capabilities. However, traditional DAB converters are insufficient for connecting multiple energy sources. Therefore, Quad-Active Bridge (QAB) DC-DC converters have gained increasing attention as a highly efficient, ultra-compact, and low-cost method for interconnecting multiple energy sources with the public power grid. By inheriting the characteristics of DAB converters—current isolation and bidirectional power flow control—QAB converters achieve higher power density and efficiency through the integration of multiple energy sources and loads. However, due to the use of multi-port high-frequency transformers, power transmission is highly coupled across the ports of a QAB converter; adjusting the phase angle between any two ports affects the power flow in all other ports. This strong coupling makes independent power control very difficult, increasing the computational burden on the control algorithm.

[0003] To eliminate the strong coupling between different ports of the QAB and achieve fast dynamic response and robustness, a control strategy combining an Extended State Observer (ESO) and Sliding Mode Control (SMC) is proposed. The core idea behind this strategy's efficient control is to utilize the ESO to observe and compensate for the system's total disturbance in real time, allowing the simplified system to be precisely controlled by a well-designed sliding mode controller. The ESO continuously receives the system's control input shift ratio and measurable output load voltage, and calculates an estimate of the total disturbance in real time. This estimated total disturbance value is fed forward into the sliding mode control law, achieving active compensation for the closed-loop control system. The disturbance is then canceled out through the efficient disturbance suppression effect of the sliding mode control. After ESO compensation, the system is "linearized" and "decoupled," and its dynamic behavior is closer to the ideal simplified model. At this point, the sliding mode controller acts on this simplified system, generating the final control signal.

[0004] To address the issues of slow dynamic response and insufficient robustness in traditional PI control, this invention fundamentally improves the dynamic response and robustness of the system by integrating a nonlinear extended state observer with superspiral sliding mode control, achieving a qualitative leap in performance. Traditional PI control works well in linearized systems, but its effectiveness is limited for QAB systems with strong nonlinearity. While this invention is slightly inferior to model predictive control in handling constraints, its analytical control law significantly reduces the computational burden and has extremely low dependence on the accuracy of system modeling. Therefore, this invention's method ensures both strong dynamic response and practical applicability in real-world applications. Summary of the Invention

[0005] This invention provides a super-spiral sliding mode active disturbance rejection control strategy based on a four-active-bridge converter, exhibiting excellent control performance. Traditional PI methods suffer from slow dynamic response and poor robustness in nonlinear systems, and struggle to eliminate the influence of strong coupling relationships between ports. This invention transforms the multi-input multi-output control system caused by the high-frequency transformer into an independent single-input single-output control system. A tracking differentiator is introduced to track the given voltage, eliminating the impact of sudden voltage changes on system stability. An extended state observer is used to estimate the total system disturbance, which is then input into a closed-loop system composed of sliding mode controllers for elimination, achieving independent control of all port voltages. Furthermore, this invention employs the super-spiral algorithm from second-order sliding mode control, fundamentally suppressing chattering while ensuring high-precision convergence of the system state within a finite time, comprehensively improving the reliability, accuracy, and efficiency of the control.

[0006] The QAB converter consists of four active H-bridges. This invention defines one active H-bridge on the primary side and three active H-bridges on the secondary side. The primary side has one port, and the secondary side has ports 2, 3, and 4 respectively. Each bridge arm contains two fully controlled switches, and the upper and lower switch signals of the same bridge arm are complementary. The transformer turns ratio is n1:n2:n3:n4. C1-C4 are the DC support capacitors for the four DC ports, and R2-R4 are the load resistors for the three secondary ports. s L is the switching frequency of the switching transistor. ij v is the equivalent inductance between ports i and j. i Let d be the port voltage referred from port i to the primary side. The inter-bridge shift ratio is defined as d. ij =d j -d i , where d ij For the shift ratio between port i and port j, this invention adopts single-phase shift control and no bridge internal shift ratio; by controlling the shift ratio relationship between each port, the energy flow between each port can be realized, and in this invention, the energy flows from the primary side to the secondary side.

[0007] In the scheme of this invention, an extended state observer is used to observe system errors, estimate uncertainties, unmodeled dynamics, coupling terms, and external disturbances in the system, and does not overly rely on the system model. The coupling between ports on the secondary side is observed as a disturbance signal to achieve dynamic decoupling control. An extended state observer is established for an nth-order system, and its mathematical description is as follows:

[0008]

[0009] The second-order nonlinear ESO state equations for the QAB system are established as follows:

[0010]

[0011] in For the unknown total disturbance of the system, b o The coefficients are the system control inputs. To ensure the observer is stable and has good performance, the gains β1, β2, and β3 need to be selected appropriately. We hope that the observer's error dynamics ( e o The eigenvalues ​​of a given point are all located in the left half-plane and have sufficiently large negative real parts to ensure fast convergence. A very popular and simple configuration method is the bandwidth method, which configures all poles in the left half-plane. o (w o >0), here w o This is called the observer bandwidth.

[0012] The technical solution of the present invention is as follows:

[0013] The multi-input multi-output system in this invention suffers from strong coupling between active H-bridges due to the high-frequency transformers used in the QAB system. This coupling makes it difficult to achieve independent control between ports, and a fault in one port significantly impacts the others. Therefore, before implementing sliding mode control, it is necessary to establish an average model and a small-signal model for the QAB to solve for the sliding mode independent control law. The modeling results are as follows:

[0014] Average model:

[0015]

[0016] Small signal model:

[0017]

[0018]

[0019]

[0020] In the above formula, g2, g3, and g4 are the average value of the inductor current at ports 2-4 over half a cycle divided by the port support capacitance value, respectively. ij For the coefficients of the state variables, F ij The coefficients of the input variables are denoted as φ. Based on the principle of active disturbance rejection control, irrelevant disturbance information caused by coupling is denoted as φ. i The formula obtained by transforming the established small-signal model is as follows:

[0021]

[0022] Where b2, b3, and b4 represent the coefficients of the control variables, and φ2, φ3, and φ4 are the total disturbances independent of this port, as shown in the following expressions:

[0023]

[0024] After modeling the QAB, in order to optimize the converter's operating performance, sliding mode control is used to suppress system disturbances and eliminate system coupling effects. At the same time, in order to eliminate the chattering problem existing in traditional sliding mode control and further improve the dynamic response capability of the system, the superhelical algorithm is introduced to design and optimize the controller. The specific implementation method is as follows:

[0025] The error equation is defined as follows:

[0026]

[0027] Among them, V refi Given the voltage value of the i-th port, V i The sliding surface is defined as follows to represent the actual voltage measurement value of the i-th port:

[0028]

[0029] After differentiating the sliding surface, the formula is as follows:

[0030]

[0031] in, φ is the control input for port i. i To address the total disturbance at port i, a superhelical algorithm is introduced. The superhelical sliding mode control method consists of two stages: the arrival stage and the sliding stage. In the arrival stage, the state is driven to a stable surface using an appropriate equivalent control law. In the sliding stage, the state slides towards a stable equilibrium point. This method has the advantage of eliminating disturbances and therefore exhibits good robustness. The formula is as follows:

[0032]

[0033] Based on the combination of the superhelical algorithm and the actual system structure, the formula is as follows:

[0034]

[0035] The control law of the system is obtained by solving for it, as shown below:

[0036]

[0037] Where α represents the coefficient of the superspiral approximation law, which aims to reach the sliding surface at a faster speed. The coefficient C determines the dynamic response capability of the system. Increasing C will enhance the dynamic response capability of the system, but may weaken the stability. Decreasing C will slow down the response process and is conducive to the stability of the system. Therefore, its value is a process that needs to be compromised according to specific needs.

[0038] After the above super-helical sliding mode self-disturbance rejection control, the voltage of each port of the system is controlled independently, the coupling effect is effectively suppressed, and more efficient and reliable control operation is achieved.

[0039] To achieve dynamic decoupling in the QAB converter, this invention proposes a four-active-bridge converter control based on a super-spiral sliding mode active disturbance rejection structure, comprising a sampling unit, an ESO observation unit, a sliding mode control unit, and a modulation unit. The sampling unit has two signal inputs: a control input shift ratio and the output load voltage. These are input into the ESO to obtain the output voltage observation value and an estimated total disturbance, calculated using the following formula:

[0040]

[0041] in, To observe the output voltage, To estimate the total disturbance, β1, β2, and β3 are the ESO gains, calculated using bandwidth w. o Adjustment;

[0042] In the sliding mode control unit, to avoid the impact caused by a step change in the reference voltage, a differential tracker is used to arrange the transient process. The voltage tracked by the differential tracker and the ESO observed voltage are input into a subtractor to obtain the error signal.

[0043]

[0044] This error is used as the input to the sliding mode controller to ensure that the system tracks the reference voltage quickly and without overshoot; an integral sliding surface is used to enhance robustness and eliminate steady-state error, as shown in the following formula:

[0045]

[0046] Where C is the sliding surface coefficient, which affects the dynamic response speed of the system. Combining the superhelical algorithm and disturbance compensation, the control quantity is obtained as follows:

[0047]

[0048] The sgn(s) function can be replaced by the sat(s) function to suppress chattering. α and η are the reaching law parameters, which control the input... Multiplying the time by half the cycle time yields the lag time, which is then input into the PWM module to generate a control signal that drives the corresponding switching transistor to turn on and off, thus achieving voltage closed-loop and power decoupling.

[0049] As can be seen from the above technical solution, the advantages of the present invention are as follows: First, the ESO, as the "intelligent sensor" of the system, can observe and compensate for the "total disturbance" in the system in real time (including model uncertainty, internal coupling and external interference), thereby decoupling the complex multi-input multi-output system into a simple independent control loop, which greatly improves the robustness and anti-interference ability of the system and reduces the dependence on the precise mathematical model; Second, the use of the superspiral algorithm fundamentally suppresses the inherent "chattering" problem of traditional sliding mode, while providing finite-time convergence characteristics, which enables it to achieve smoother, faster and more precise control while maintaining strong robustness. Attached Figure Description

[0050] Figure 1 This is the main topology diagram of the QAB converter.

[0051] Figure 2 This is a waveform diagram of the output voltage and inductor current at each port of the QAB converter under PI closed-loop control and load disturbance at port 2.

[0052] Figure 3 The waveforms of the output voltage and inductor current at each port under load disturbance at port 2 and port 1 of the QAB converter during super-spiral sliding mode control are shown.

[0053] Figure 4 This is a waveform diagram of the output voltage and inductor current at the ports of a QAB converter under PI closed-loop control when the given voltage at the two ports changes abruptly.

[0054] Figure 5 The waveforms of the port output voltage and inductor current of the QAB converter under the sudden change of the given voltage at the two ports during super-spiral sliding mode control are shown.

[0055] Figure 6 This is the closed-loop control system of the present invention. Detailed Implementation

[0056] The present invention will be further described below with reference to the accompanying drawings, but this should not be construed as limiting the scope of protection of the present invention.

[0057] like Figure 1As shown, the QAB converter consists of four active H-bridges, four high-frequency inductors, and one high-frequency isolation transformer. The primary side has one active H-bridge, and the secondary side has three active H-bridges. Each bridge arm contains two fully controlled switches. i For high-frequency inductors, N i C1 is the number of turns at each port of the transformer; C2, C3, and C4 are the input capacitors; V1 is the input voltage; R... i This is the output-side resistance.

[0058] Figure 2 The figure shows the output voltage waveforms of the QAB converter under 2-port load disturbance during PI closed-loop control. It can be seen from the figure that when the 2-port load decreases from 10Ω to 5Ω, the voltage on the 2-port side drops by about 3.9V, and the voltage on other secondary ports drops by about 1.2V. It can be seen that the system dynamic response is slow under traditional PI control and there is still a large coupling effect between the ports, resulting in poor robustness. Figure 4 The output voltage setpoint for the two ports is sequentially stepped from 50V to 60V, and then to 40V. Under traditional PI control, the system settling time from 60V to 40V is 6.9ms, the response is relatively slow, and the time required to recover to a stable state is long. Based on this, this invention uses an Extended State Observer (ESO) to observe the coupling between the ports on the secondary side as a disturbance signal, thereby achieving dynamic decoupling control. The ESO is used to estimate uncertainties, unmodeled dynamics, coupling terms, and external disturbances in the system, without overly relying on the system model. An extended state observer is established for an nth-order system, the mathematical description of which is as follows:

[0059]

[0060] The second-order ESO state equations for the QAB system are as follows:

[0061]

[0062] in For the unknown total disturbance of the system, b o The coefficients are the system control inputs. To ensure the observer is stable and has good performance, the gains β1, β2, and β3 need to be selected appropriately. We hope that the observer's error dynamics ( e o The eigenvalues ​​of a given point are all located in the left half-plane and have sufficiently large negative real parts to ensure fast convergence. A very popular and simple configuration method is the bandwidth method, which configures all poles in the left half-plane. o (w o >0), here w o This is called the observer bandwidth.

[0063] Figure 3The graph shows the output voltage waveforms of the QAB converter under a 2-port load disturbance during super-slip sliding mode control. It can be seen from the graph that when the 2-port load decreases from 10Ω to 5Ω, the voltage at port 2 drops by approximately 0.43V, while the voltage at other secondary ports drops by approximately 0.13V. Figure 5 To achieve a 2-port output voltage setpoint that sequentially jumps from 50V to 60V and then to 40V under superspiral sliding mode control, the settling time from 60V to 40V is reduced to 3.4ms, approximately 50% less than that of traditional PI control. This demonstrates faster dynamic adjustment speed and stronger command tracking capability. Compared to PI, superspiral sliding mode control significantly improves dynamic response and coupling suppression, exhibiting excellent control capabilities for multi-input multi-output systems like QABs. The superspiral algorithm used can suppress chattering issues present in traditional sliding mode controllers. The small-signal modeling of the QAB is as follows:

[0064]

[0065] The sliding surface is created as follows:

[0066]

[0067] Based on the combination of the superhelical algorithm and the actual system structure, the formula is as follows:

[0068]

[0069] The control law of the system is obtained by solving for it, as shown below:

[0070]

[0071] Where α represents the coefficient of the superspiral approximation law, which aims to reach the sliding surface at a faster speed. The coefficient C determines the dynamic response capability of the system. Increasing C will enhance the dynamic response capability of the system, but may weaken the stability. Decreasing C will slow down the response process and is conducive to the stability of the system. Therefore, its value is a process that needs to be compromised according to specific needs.

[0072] like Figure 4 As shown, this invention employs a four-active-bridge converter control based on a super-spiral sliding mode self-disturbance rejection structure. This closed-loop control system includes the following sequential steps:

[0073] 1) Based on the small-signal model of the QAB converter, coupling terms between ports, external load disturbances, parameter uncertainties, etc., are uniformly regarded as the "total disturbance" φ. i The system model can be rewritten as follows:

[0074]

[0075] Then, a second-order ESO is designed to expand the total perturbation into new state variables, as shown below:

[0076]

[0077] 2) A differential tracker is used to arrange the transient process. The voltage tracked by the differential tracker and the voltage observed by the ESO are input into a subtractor to obtain the error signal:

[0078]

[0079] 3) An integral sliding surface is used to enhance robustness and eliminate steady-state error. The formula is as follows:

[0080]

[0081] Combining the superhelical algorithm and disturbance compensation, the control variables are obtained as follows:

[0082]

[0083] control quantity Multiplying the time by half the cycle time yields the lag time, which is then input into the PWM module to generate a control signal that drives the corresponding switching transistor to turn on and off, thus achieving voltage closed-loop and power decoupling.

Claims

1. A super-spiral sliding mode active disturbance rejection control strategy based on a four-active-bridge converter, characterized in that, The multi-input multi-output control system caused by the high-frequency transformer is transformed into an independent single-input single-output control system. A tracking differentiator is introduced to track the given voltage, eliminating the impact of sudden changes in the given voltage on the system's operational stability. Combined with an extended state observer to estimate the total system disturbance, the disturbance estimate is input into a closed-loop system composed of sliding mode controllers for elimination, achieving independent control of the voltage at all ports. In addition, this invention employs the super-spiral algorithm in the second-order sliding mode control method, which fundamentally suppresses chattering while ensuring high-precision convergence of the system state within a finite time, comprehensively improving the reliability, accuracy, and efficiency of the control.

2. The quad-active bridge (QAB) converter according to claim 1, characterized in that, The QAB converter consists of four active H-bridges. This invention defines one active H-bridge on the primary side and three active H-bridges on the secondary side. The primary side has one port, and the secondary side has ports 2, 3, and 4 respectively. Each bridge arm contains two fully controlled switches, and the upper and lower switch signals of the same bridge arm are complementary. The transformer turns ratio is n1:n2:n3:n4. C1-C4 are the DC support capacitors for the four DC ports, and R2-R4 are the load resistors for the three secondary ports. s L is the switching frequency of the switching transistor. ij v is the equivalent inductance between ports i and j. i Let d be the port voltage referred from port i to the primary side. The inter-bridge shift ratio is defined as d. ij =d j -d i , where d ij Compared to the shift ratio between port i and port j, this invention employs single-phase shift control and eliminates the bridge-inward shift ratio; in this invention, the energy flows from the primary side to the secondary side.

3. The multiple-input multiple-output system according to claim 1, characterized in that, Because the high-frequency transformers used in the QAB system cause strong coupling between the active H-bridges, this coupling makes it difficult to achieve independent control between the ports. Furthermore, a fault in one port has a significant impact on the other ports. Therefore, before implementing sliding mode control of the system, it is necessary to establish an average model and a small-signal model for the QAB to solve for the sliding mode independent control law. The modeling results are as follows: Average model: Small signal model: In the above formula, g2, g3, and g4 are the average value of the inductor current at ports 2-4 over half a cycle divided by the port support capacitance value, respectively. ij For the coefficients of the state variables, F ij The coefficients of the input variables; Based on the principle of active disturbance rejection control, coupling information in the system that is unrelated to this port is treated as a disturbance and denoted as φ. i The formula obtained by transforming the small-signal model is as follows: Where b2, b3, and b4 represent the coefficients of the control variables, and φ2, φ3, and φ4 are the total disturbances independent of this port, as shown in the following expressions:

4. The superspiral sliding mode self-disturbance rejection control according to claim 1, characterized in that, After modeling the QAB, in order to optimize the converter's operating performance, sliding mode control is used to suppress system disturbances and eliminate system coupling effects, thereby improving the system's dynamic response capability. At the same time, the superhelical algorithm is used to suppress chattering problems existing in traditional sliding mode controllers and improve control accuracy. The specific implementation method is as follows: The error equation is defined as follows: Among them, V refi Given the voltage value of the i-th port, V i The sliding surface is defined as follows to represent the actual voltage measurement value of the i-th port: After differentiating the sliding surface, the formula is as follows: in, φ is the control input for port i. i To address the total disturbance at port i, a superhelical algorithm is introduced. The superhelical sliding mode control method consists of two phases: the arrival phase and the sliding phase. In the arrival phase, the state is driven to a stable surface using an appropriate equivalent control law. In the sliding phase, the state slides towards a stable equilibrium point. This method has the advantage of eliminating disturbances and therefore exhibits good robustness. The formula is as follows: Based on the combination of the superhelical algorithm and the actual system structure, the formula is as follows: The control law of the system is obtained by solving for it, as shown below: Where α represents the coefficient of the superspiral approximation law, which aims to reach the sliding surface at a faster speed. The coefficient C determines the dynamic response capability of the system. Increasing C will enhance the dynamic response capability of the system, but may weaken the stability. Decreasing C will slow down the response process and is conducive to the stability of the system. Therefore, its value is a process that needs to be compromised according to specific needs.

5. The φ as described in claim 4 i To mitigate the observation error of the extended state observer, dynamic decoupling control is achieved by using an ESO to treat the coupling observations between the ports of the secondary side as disturbance signals; its key feature is that... To estimate uncertainties, unmodeled dynamics, coupling terms, and external disturbances in the system without over-relying on the system model, an extended state observer is established for an nth-order system, mathematically described as follows: The second-order ESO state equations for the QAB system are as follows: in For the unknown total disturbance of the system, b o The coefficients are the system control inputs. To ensure the observer is stable and has good performance, the gains β1, β2, and β3 need to be selected appropriately. We hope that the observer's error dynamics ( e o The eigenvalues ​​of a given point are all located in the left half-plane and have sufficiently large negative real parts to ensure fast convergence. A very popular and simple configuration method is the bandwidth method, which configures all poles in the left half-plane. o (w o >0), here w o This is called the observer bandwidth.

6. The control of the four active bridge converter based on the superspiral sliding mode self-disturbance rejection structure according to claim 1, characterized in that, The control process specifically includes the following steps: Step 1: First, based on the small-signal model of the QAB converter, the coupling terms between ports, external load disturbances, parameter uncertainties, etc., are all considered as the "total disturbance" φ. i The system model can be rewritten as follows: Where, φ i This includes all coupling terms and external disturbances caused by non-port control variables; then a second-order ESO is designed to expand the total disturbance into new state variables, as shown below: in, To observe the output voltage, To estimate the total disturbance, β1, β2, and β3 are the ESO gains, calculated using bandwidth w. o Adjustment; Step 2: To avoid the impact caused by a step change in the reference voltage, a differential tracker is used to arrange the transient process. The voltage tracked by the differential tracker and the voltage observed by the ESO are input into a subtractor to obtain the error signal. This error is used as input to the sliding mode controller to ensure that the system tracks the reference voltage quickly and without overshoot. Step 3: Employ an integral sliding surface to enhance robustness and eliminate steady-state error, as shown in the following formula: Where C is the sliding surface coefficient, which affects the dynamic response speed of the system. Combining the superhelical algorithm and disturbance compensation, the control quantity is obtained as follows: The sgn(s) function can be replaced by the sat(s) function to suppress chattering. α and η are the reaching law parameters, which control the input... Multiplying the time by half the cycle time yields the lag time, which is then input into the PWM module to generate a control signal that drives the corresponding switching transistor to turn on and off, thus achieving voltage closed-loop and power decoupling.

Images