An AI self-tuning DC converter system based on a coupled double-voltage resonance network

Abstract

The application discloses an AI self-tuning DC converter system based on a coupling voltage-doubler resonant network, and belongs to the technical field of power electronic converters. The AI self-tuning DC converter system comprises an input filtering and switching circuit, a charging and discharging circuit, a voltage-doubler filtering output circuit and a deep learning control subsystem. The turns ratio of a three-winding coupling inductor and the duty cycle D of a switching tube are cooperatively adjusted through the voltage-doubling effect of the voltage-doubler filtering output circuit and the charging and discharging characteristics of the three-winding coupling inductor, so that the DC converter realizes high-gain and high-flexibility voltage boosting performance. The topology realizes soft switching through resonance formed by leakage inductance and capacitance, and significantly improves working efficiency. Through deep learning and AI training, the specific parameters of elements in the tuning circuit and the parasitic parameters are determined, and the AI self-tuning PID parameter mode is adopted to simplify the control mode and realize closed-loop control of the system.

Description

Technical Field

[0001] This invention relates to an AI self-tuning DC-DC converter system based on a coupled voltage doubler resonant network, belonging to the field of power electronic converter technology. Background Technology

[0002] With the large-scale deployment of renewable energy sources such as solar and wind power, and the development of electrification and intelligentization on the end-user electricity side, higher demands are being placed on the efficiency, reliability, and flexibility of power supply. Against this backdrop, DC microgrid technology, which can efficiently integrate distributed energy resources and flexibly allocate power, has become an important development direction for modern energy systems.

[0003] As the core power electronic device in a DC microgrid, the DC-DC converter directly impacts the system's efficiency, power density, and operational stability, enabling voltage level conversion, energy transfer, and management. Given the inherent intermittent and fluctuating output characteristics of renewable energy sources, DC-DC converters require high voltage gain to effectively boost the low-voltage DC power generated by photovoltaic panels to the bus voltage level. The increasing miniaturization, lightweight design, and high efficiency of end-user equipment necessitates breakthroughs in improving power density and conversion efficiency for DC-DC converters. To achieve this, the industry generally favors wide-bandgap semiconductor devices, reducing the size of passive components by increasing the switching frequency.

[0004] The literature “ZHENG Yifei, BROWN B, XIE Wenhao, et al. High step-up DC-DC converter with zero voltage switching and low input current ripple[J]. IEEE Transactions on Power Electronics, 2020, 35(9): 9416-9429.” uses synchronous rectifier switching transistors to replace diodes to achieve zero voltage switching (ZVS) of the switching transistors. At the same time, the zero DC bias of the coupled inductor can obtain a smaller magnetic size and lower core loss. However, this technology has the problems of low voltage gain and large voltage stress on the output diode. The literature “ZAOSKOUFIS K, TATAKIS EC. Isolated ZVS-ZCS DC-DC high step-up converter with low-ripple input current[J]. IEEE Journal of Emerging and Selected Topics in Industrial Electronics, 2021, 2(4): 464-480. "The soft-switching operation of all semiconductor devices is achieved, but the converter uses an isolation transformer, which reduces efficiency and separates the ground between the input and output; the literature "Alavi P, Mohseni P, Babaei E, et al. Anultra-high step-up DC–DC converter with extendable voltage gain and soft-switching capability[J]. IEEE Transactions on Industrial Electronics, 2020, 67(11): 9238-9250." proposes an scalable high-gain soft-switching converter, which achieves soft switching of the switching transistor by designing the magnitude of the excitation current, and achieves soft switching of the diode by controlling the current change slope through leakage inductance; the literature "Heidari M, Esteki M, Khajehoddin SA, et al. A high voltage gain ZVT quasi-Z-sourceconverter with reduced voltage stress[J].IEEE Transactions on Power Electronics, 2022, 37(11): 13696-13710. This paper proposes a converter based on a quasi-Z-source network, integrating switched capacitors and coupled inductors. It utilizes an auxiliary switch to form a low-impedance resonant circuit, rapidly discharging the parasitic capacitance of the main switch. The main switch is then turned on early via its body diode, achieving soft switching. The diode and auxiliary switch achieve soft switching through the resonance of the leakage inductance of the coupled inductor and the circuit capacitance. However, this converter uses a large number of components.

[0005] In summary, how to comprehensively achieve high boost capability, high efficiency, high power density, and high reliability of DC-DC converters in complex and demanding application scenarios has become a core technical challenge that urgently needs to be addressed by those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide an AI self-tuning DC-DC converter system based on a coupled voltage doubler resonant network. Through innovation in transformer structure and parameter optimization, the overall performance of the converter is significantly improved, featuring high voltage gain, soft switching and high efficiency.

[0007] To achieve the above objectives / to solve the above technical problems, the present invention is implemented using the following technical solution:

[0008] This invention provides an AI self-tuning DC-DC converter system based on a coupled voltage doubler resonant network, including an input filtering and switching circuit, a charging and discharging circuit, a voltage doubler filter output circuit, and a deep learning control subsystem;

[0009] The input filtering and switching circuit includes a DC regulated input power supply, an input filter inductor, and a semiconductor switching transistor connected in series.

[0010] The charging and discharging circuit includes a first energy storage capacitor, a second coupling inductor, a stray inductor, and a first coupling inductor connected in series. The starting end of the first energy storage capacitor is connected to a semiconductor switch. Magnetizing inductors are connected in parallel on both sides of the first coupling inductor. The anode of the first passive switch is connected to the starting end of the first input energy storage capacitor, and the cathode of the first passive switch is connected to the same-name end of the first coupling inductor.

[0011] The voltage multiplier filter output circuit includes a third coupling inductor, a second energy storage capacitor, a second passive switch, an output diode, a first output filter capacitor, and a second output filter capacitor. The same-name terminal of the third coupling inductor is connected to the same-name terminal of the first coupling inductor, and the opposite-name terminal of the third coupling inductor is connected to one end of the second energy storage capacitor. The other end of the second energy storage capacitor is connected to the cathode of the second passive switch and the anode of the output diode, respectively. The cathode of the output diode is connected to one end of the second output filter capacitor. The same-name terminal of the third coupling inductor, the anode of the second passive switch, and the other end of the second output filter capacitor are respectively connected to one end of the first output filter capacitor. The other end of the first output filter capacitor is connected to the negative terminal of the DC regulated input power supply.

[0012] The input terminal of the deep learning control subsystem is connected to a resistor, which is connected in parallel across the first and second output filter capacitors. The output terminal of the deep learning control subsystem is connected to a semiconductor switching transistor.

[0013] Furthermore, the deep learning control subsystem includes an output voltage sampling module, a reference voltage setting module, an error comparator, an integrator, a differentiator, an AI control operator module, a deep learning controller, and a modulator. The output voltage sampling module is used to acquire the resistor voltage, the reference voltage setting module is used to set the desired voltage, the error comparator is used to compare the resistor voltage and the desired voltage to obtain a voltage error signal, and the AI ​​control operator module is used to output a first reference control parameter K based on the voltage error signal. d0 Second reference control parameter K i0 Third reference control parameter K p0 The deep learning controller is used to control the first baseline parameter K. d0 Second reference control parameter K i0 Third reference control parameter K p0 Output the first correction amount ΔK d0 Second correction amount ΔK i0 The third correction amount ΔK p0 The integrator and differentiator are used to integrate and differentiate the voltage error signal, respectively, and the modulator is used to control the voltage error signal according to the first reference parameter K. d0 Second reference control parameter K i0 Third reference control parameter K p0 First correction amount ΔK d0 Second correction amount ΔK i0 The third correction amount ΔK p0 The voltage error signal after differentiation and the voltage error signal after integration are output as PWM signals to control the semiconductor switching transistors.

[0014] Furthermore, the first reference control parameter K d0 Second reference control parameter Ki0 Third reference control parameter K p0 With the first correction amount ΔK d0 Second correction amount ΔK i0 The third correction amount ΔK p0 By successively subtracting, the corrected first control parameter K is obtained. d Second control parameter K i The third control parameter K p The voltage error signal is integrated using an integrator and then compared with the second control parameter K. i Multiplying them together yields the second intermediate quantity K. i The voltage error signal is differentiated using a differentiator and then compared with the first control parameter K. d Multiplying them together yields the first intermediate quantity K. d ', the third control parameter K p The third intermediate quantity K is obtained by multiplying it by the voltage error signal. p ', will K d '、K i '、K p Input a PWM modulator, and use the PWM modulator to output a PWM signal for controlling the semiconductor switching transistor.

[0015] Furthermore, the first, second, and third coupled inductors are three-winding coupled inductors with a common magnetic core. The number of turns of the coils of the first, second, and third coupled inductors are R1, R2, and R3, respectively. n1 = R2 / R1 and n2 = R3 / R1 are defined, where n1 and n2 are real numbers not equal to 1.

[0016] Furthermore, the semiconductor switch is an insulated gate bipolar transistor or a metal-oxide-semiconductor field-effect transistor.

[0017] Furthermore, the semiconductor switch also includes a parasitic diode, the anode of which is connected to the source of the semiconductor switch, and the cathode of which is connected to the gate of the semiconductor switch. The drain of the semiconductor switch is connected to the output of the deep learning control subsystem, and the switching on and off of the parasitic diode is controlled by receiving the PWM signal output by the deep learning control subsystem.

[0018] Furthermore, the voltage gain of the system is controlled by coordinating the turns ratio of the first, second, and third coupled inductors and the duty cycle of the semiconductor switch.

[0019] Furthermore, the DC-DC converter system output voltage V o The expression for the output gain B is as follows:

[0020] ;

[0021] ;

[0022] in, The voltage of the DC regulated input power supply is n1 = R2 / R1, n2 = R3 / R1, R1, R2 and R3 are the number of turns of the first coupling inductor CI1, the second coupling inductor CI2 and the third coupling inductor CI3 respectively, and D is the duty cycle of the semiconductor switch PST1.

[0023] Furthermore, by adjusting the parameters of the magnetizing inductor, stray inductor, first energy storage capacitor, buffer capacitor of the semiconductor switch, and first output filter capacitor, the semiconductor switch is turned on under zero voltage conditions under the action of the drive signal; and through the resonance effect between the first coupling inductor, second coupling inductor, third coupling inductor, and each capacitor, when the semiconductor switch is turned on, the first passive switch, second active switch, and output switch are turned off with zero voltage and zero current.

[0024] Furthermore, in the DC-DC converter system, the first energy storage capacitor, the first coupling inductor, and the first passive switch constitute the first clamping unit; the third coupling inductor, the second passive switch, and the second energy storage capacitor constitute the second clamping unit; and the third coupling inductor, the second energy storage capacitor, the output diode, and the second output filter capacitor constitute the third clamping unit.

[0025] The three clamping units interact in steady state. The first clamping unit stores energy in the first energy storage capacitor and controls the voltage stress across the semiconductor switch. The second clamping unit and the third clamping unit absorb the leakage inductance energy of the third coupling inductor and limit the voltage stress on the second passive switch and the output diode.

[0026] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0027] This invention proposes an AI self-tuning DC-DC converter system based on a coupled voltage doubler resonant network. By optimizing the energy transfer path, the overall energy efficiency is improved. Through the voltage boosting effect of the voltage doubler unit structure, combined with the charging and discharging characteristics of the three coupled inductors, the turns ratio of the three-winding coupled inductors and the duty cycle D of the switching transistors are synergistically adjusted, enabling the DC-DC converter to achieve high gain and high flexibility in voltage boosting performance, improving voltage gain and reducing device stress. This invention can also achieve soft switching through the resonance formed by leakage inductance and capacitance in the circuit, improving converter efficiency. Finally, this invention introduces deep learning and AI training technology into the DC-DC converter system. Through deep learning, the specific parameters of the components in the tuning circuit are controlled, and parasitic parameters are determined. A PWM signal is output to dynamically adjust the switching of the semiconductor switching transistors, achieving closed-loop control of the system. Attached Figure Description

[0028] Figure 1 The figure shown is a schematic diagram of an AI self-tuning DC-DC converter system based on a coupled voltage doubler resonant network provided by the present invention.

[0029] Figure 2 The diagram shown is a schematic of a deep learning controller in an embodiment of the present invention;

[0030] Figure 3 The diagram shown is an equivalent circuit diagram of the converter in a stable operating cycle [t0, t5] in an embodiment of the present invention.

[0031] Figure 4 The diagram shown is an equivalent circuit diagram of the semiconductor switch PST1 of the converter in an embodiment of the present invention, showing the conduction times [t1, t2].

[0032] Figure 5 The diagram shown is an equivalent circuit diagram of the semiconductor switch PST1 of the converter in an embodiment of the present invention at the turn-off times [t4, t5].

[0033] Figure 6 The diagram shown is a schematic diagram illustrating the relationship between voltage stress and duty cycle of the active device in the converter in an embodiment of the present invention.

[0034] Figure 7 The diagram shown illustrates the relationship between converter voltage gain and transformer turns ratio and duty cycle in an embodiment of the present invention.

[0035] Figure 8 The figure shown is a schematic diagram of a closed-loop simulation experiment of a transformer under the control method of using AI self-tuning parameters in an embodiment of the present invention.

[0036] In the diagram, 1-input filtering and switching circuit, 2-charging and discharging circuit, 3-voltage multiplier filter output circuit, and 4-deep learning control subsystem. Detailed Implementation

[0037] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0038] Example 1

[0039] This embodiment introduces an AI self-tuning DC-DC converter system based on a coupled voltage doubler resonant network, such as... Figure 1 As shown, it mainly includes input filtering and switching circuit 1, charging and discharging circuit 2, voltage multiplier filter output circuit 3, and deep learning control subsystem 4.

[0040] The input filtering and switching circuit includes a DC regulated input power supply V. s Input filter inductor L in And semiconductor switching transistor PST1, DC regulated input power supply V s The positive terminal is connected to the input filter inductor L. in One end of the semiconductor switch PST1 is connected to the second end of the input filter inductor L. in At the other end, the third terminal of the semiconductor switch PST1 is connected to the DC regulated input power supply V. s The negative terminal of the transistor is connected to GND, and the first terminal of the semiconductor switch PST1 is connected to the deep learning control subsystem. The semiconductor switch PST1 can be implemented using an insulated-gate bipolar transistor (IGBT) or a metal-oxide-semiconductor field-effect transistor (MOSFET). When using an IGBT, its first, second, and third terminals correspond to the collector, gate, and emitter, respectively; when using a MOSFET, its first, second, and third terminals correspond to the drain, gate, and source, respectively.

[0041] The charging and discharging circuit includes a first energy storage capacitor C1, a second coupling inductor CI2, and a secondary normalized stray inductor L connected in series. k The first coupling inductor CI1 and the first energy storage capacitor C1 are connected at their starting ends to the second end of the semiconductor switch PST1. The two sides of the first coupling inductor CI1 are connected in parallel to the excitation inductor L, which is calculated from the secondary side. m The anode of the first passive switch D1 is connected to the starting terminal of the first input energy storage capacitor C1, and the cathode of the first passive switch D1 is connected to the same terminal of the first coupling inductor CI1. This invention constructs a charging and discharging circuit with stray inductance modeling, which can optimize the energy transfer path and improve overall energy efficiency.

[0042] The voltage multiplier filter output circuit includes a third coupling inductor CI3, a second energy storage capacitor C2, a second passive switch D2, and an output diode D. o First output filter capacitor C o1 Second output filter capacitor C o2 The same-name terminal of the third coupling inductor CI3 is connected to the same-name terminal of the first coupling inductor CI1, and the opposite-name terminal of the third coupling inductor CI3 is connected to one end of the second energy storage capacitor C2. The other end of the second energy storage capacitor C2 is connected to the cathode of the second passive switch D2 and the output diode D1, respectively. o The anode, output diode D o The cathode is connected to the second output filter capacitor C. o2 One end, the same-name terminal of the third coupling inductor CI3, the anode of the second passive switch D2, and the second output filter capacitor C o2 The other end is connected to the first output filter capacitor C. o1 One end, the first output filter capacitor Co1 The other end is connected to a DC regulated input power supply V. s The negative terminal. The voltage doubler filter output circuit uses the synergistic charging and discharging effect of the energy storage capacitor and the coupling inductor to superimpose the voltages of different circuit nodes onto the output terminal, thereby achieving a high voltage gain in a single structure and overcoming the limitations of traditional single inductor or capacitor boosting methods in terms of gain capability.

[0043] In addition, the system also has a first output filter capacitor C o1 Second output filter capacitor C o2 A resistor R is connected in parallel across the circuit. The output voltage V can be obtained by measuring the voltage across resistor R. o .

[0044] The deep learning control subsystem includes an output voltage sampling module, a reference voltage setting module, an error comparator, an integrator, a differentiator, an AI control operator module, a deep learning controller, and a modulator. The input of the output voltage sampling module is connected to a resistor R to acquire the voltage across the resistor; the reference voltage setting module is used to set the desired voltage V. ref The error comparator CMP1 is used to compare the resistor voltage and the desired voltage to obtain the voltage error signal e(t); the integrator and differentiator are used to perform integration and differentiation operations on the voltage error signal e(t), respectively; the AI ​​control operator module is used to output the reference control parameters based on the voltage error signal e(t) and its rate of change du / dt; the deep learning controller is used to output the correction amount of the reference control parameters through the deep learning algorithm; the output of the modulator is connected to the semiconductor switch PST1, and is used to dynamically adjust the switching strategy of the semiconductor switch PST1 through the output PWM signal.

[0045] In embodiments of the present invention, the deep learning controller is as follows: Figure 2 As shown, the deep learning controller includes a feature extraction layer, a deep learning model, and a parameter output layer. The deep learning model is primarily a CNN+LSTM model. In actual operation, the key inputs to the deep learning controller include the error e(t) between the output voltage and the reference voltage, the rate of change of the error signal over time du / dt, and the circuit output voltage V0.

[0046] In the feature extraction layer, e(t), du / dt, and V0 are first preprocessed, mainly including normalization, filtering, and construction of historical data sequences. Normalization eliminates the amplitude influence between different inputs, standardizing them into a uniform interval. Filtering removes signal noise and improves data stability. The feature extraction layer also combines e(t), du / dt, and V0 within the current time and a certain historical window into a multi-dimensional feature sequence, enhancing the model's ability to capture dynamic processes and trend information. For e(t) and du / dt data, the feature dimensions can be expanded using sliding windows, statistical features (such as mean and variance), and recent peak values, enabling the deep learning model to better learn the steady-state and dynamic behavior of the system.

[0047] In deep learning models, features extracted by the feature extraction layer are automatically analyzed and learned. The deep learning model first extracts local temporal and correlation features through convolutional layers, and further analyzes long-term dependencies and dynamic changes using structures such as LSTM. The model continuously optimizes parameters during training, adaptively mining key patterns affecting system performance from the input features. Finally, multiple correction values ​​(ΔK) are output at the parameter output layer. p ΔK i ΔK d ).

[0048] In the deep learning control subsystem, this invention introduces deep learning and AI training, based on the actual output voltage V. o and desired voltage V ref The system automatically learns and trains to tune the specific parameters of components in the circuit and determine parasitic parameters. By using AI to self-tune PID parameters, the control method is simplified, achieving closed-loop control of the system. Regardless of disturbances or load changes encountered by the system, the deep learning model can dynamically adjust the control strategy, enabling the output voltage to stably and efficiently track the target reference value, significantly improving the system's response speed, steady-state accuracy, and robustness.

[0049] The working principle of the system of this invention is as follows:

[0050] DC regulated input power supply By input filter inductor L in To power the system, the semiconductor switch PST1 is periodically switched on and off under the control of a drive signal. When the switch is on, the input energy and the energy stored in the first energy storage capacitor C1 are transferred to the coupled inductor network through a charging and discharging circuit. The charging and discharging circuit includes the secondary-side normalized stray inductance L used for modeling and utilizing leakage inductance energy. kThis inductor works in conjunction with the first coupling inductor CI1 and the second coupling inductor CI2 to achieve efficient energy transfer and storage. When the switch is turned off, the energy stored in the coupling inductor network and the second energy storage capacitor C2 is released to the voltage doubler filter output circuit. The voltage doubler circuit is connected to the third coupling inductor CI3, the second energy storage capacitor C2, and the output filter capacitor C. o1 C o2 The synergistic effect of the components superimposes voltages from different nodes at the output, thereby achieving high voltage gain. The output voltage is sampled by a parallel resistor and then fed back to the deep learning control subsystem to complete the real-time self-tuning and optimized operation of the system.

[0051] In the deep learning control subsystem, the desired voltage V is set through the reference voltage setting module. ref The output voltage sampling module collects the resistor voltage V. o Using error comparator CMP1 to measure the resistor voltage V o and desired voltage V ref The voltage error signal is obtained through comparison and then input to the AI ​​control operator module. The AI ​​control operator module outputs the first reference control parameter K. d0 Second reference control parameter K i0 Third reference control parameter K p0 The deep learning controller, based on the first reference control parameter K, then... d0 Second reference control parameter K i0 Third reference control parameter K p0 Dynamically adjust and output the first correction amount ΔK d0 Second correction amount ΔK i0 The third correction amount ΔK p0 The first reference control parameter K d0 Second reference control parameter K i0 Third reference control parameter K p0 With the first correction amount ΔK d0 Second correction amount ΔK i0 The third correction amount ΔK p0 The corrected first control parameter K is obtained by successively subtracting the values. d Second control parameter K i The third control parameter K p The voltage error signal is integrated using an integrator and then compared with K. i Multiplying them together yields the second intermediate quantity K. i '( Figure 1 (Not marked), the voltage error signal is differentiated using a differentiator and then compared with K. d Multiplying them together yields the first intermediate quantity K. d ' ( Figure 1 (Not marked), Kp The third intermediate quantity K is obtained by multiplying it by the voltage error signal. p ' ( Figure 1 (Not marked), K d '、K i '、K p Input a PWM modulator, and use the PWM modulator to output a PWM signal for controlling the semiconductor switching transistor.

[0052] The system of this invention includes a three-winding coupled inductor, which consists of a first coupled inductor CI1, a second coupled inductor CI2, and a third coupled inductor CI3. This three-winding coupled inductor serves as the main energy transfer path of the converter, enabling high voltage gain. The three-winding coupled inductor is equivalent to an ideal transformer model in the circuit, with its primary winding containing a magnetizing inductance L. m Each winding has a corresponding leakage inductance L. 1k L 2k With L 3k The topological graph is normalized to the original edge, denoted as L. k The number of turns of the first coupled inductor CI1, the second coupled inductor CI2, and the third coupled inductor CI3 are R1, R2, and R3, respectively, and the turns ratio is R1:R2:R3. The turns ratio can also be expressed as 1:n1:n2, where n1=R2 / R1 and n2=R3 / R1, and n1 and n2 are real numbers not equal to 1.

[0053] By coordinating the adjustment of the duty cycle D of n1, n2 and semiconductor switch PST1, a higher voltage gain can be obtained under conditions of lower turns ratio and duty cycle, thereby achieving flexible and wide adjustment of the boost range of the converter.

[0054] The control terminal of the semiconductor switch PST1 is connected to an external controller to receive control signals and regulate its on / off state. The external controller includes, but is not limited to, an STM32 series microcontroller or a TMS320 series digital signal processor, used to generate and output control signals. In this embodiment of the invention, the external controller is a deep learning controller.

[0055] exist Figure 1In this circuit, the semiconductor switch PTS1 is an enhancement-mode N-channel metal-oxide-semiconductor field-effect transistor, also known as an N-channel MOSFET. The drain and source of PTS1 receive the PWM output from the modulator of the deep learning controller. PTS1 also includes a parasitic diode. The anode of the parasitic diode is connected to the source of PTS1, and the cathode is connected to the gate of PTS1. This parasitic diode forms a freewheeling loop in the circuit, providing a necessary current discharge path for energy storage components such as inductors, while also achieving voltage clamping to prevent harmful voltage spikes when the switch is turned off. The drain of PTS1 can be connected to the drive signal of an external driver, thereby controlling the conduction and turn-off of the parasitic diode.

[0056] In this embodiment of the invention, the main control chip of the modulator is TMS320VC5502PGF300; the driving modulation method is unipolar pulse width modulation, and a PID control method combined with AI operators is adopted.

[0057] This invention uses an input filter inductor L in The first energy storage capacitor C1, the first passive switch D1, and the primary side CI1 and secondary side CI2 of the three-winding coupled inductor can maintain continuous input current and have low input current ripple, thereby suppressing current surges to the preceding circuit.

[0058] This invention achieves this by rationally configuring the excitation inductor L m Stray inductance L k The first energy storage capacitor C1, and the buffer capacitor C of the semiconductor switching transistor PST1. s1 and the first output filter capacitor C o1 The parameters are configured such that semiconductor switch PST1 can be turned on under zero-voltage conditions when the drive signal arrives. Utilizing the resonance between the three-winding coupled inductor and the capacitors in the circuit, during the diode's conduction period, the passive switches D1 and D2, as well as the output switch D... o This invention achieves zero-voltage, zero-current turn-off, thus enabling fully soft-switching operation. It significantly reduces the turn-on and turn-off losses of switching devices, thereby improving the overall efficiency of the converter.

[0059] The system of this invention includes multiple clamping structures. Specifically, the first clamping unit consists of a first energy storage capacitor C1, a first coupling inductor CI1, and a first passive switch D1; the second clamping unit consists of a third coupling inductor CI3, a second passive switch D2, and a second energy storage capacitor C2; the third clamping unit consists of a third coupling inductor CI3, a second energy storage capacitor C2, and an output diode D1. o and the second output filter capacitor C o2Composition. Three clamping units interact in steady state. The first clamping unit stores energy in the first energy storage capacitor C1, controlling the voltage stress across the semiconductor switch PST1. The second and third clamping units absorb the leakage inductance energy of the third coupling inductor CI3, limiting the second passive switch D2 and the output diode D. o Voltage stress on.

[0060] Figure 3 The diagram illustrates key waveforms of each component during a steady-state operating cycle [t0, t5] according to an embodiment of the present invention, specifically including the control signal waveform received by semiconductor switch PST1. The voltage and current waveforms of the first passive switch D1 ( and The voltage and current waveforms of the second passive switch D2 ( and Output diode D o Voltage and current waveforms ( and Magnetizing inductance L m Current waveform ( ), and stray inductance L k Current waveform ( ).

[0061] according to Figure 3 It can be seen that there are five operating modes within a steady-state operating cycle. These five operating modes are divided by time t1-t5 and are denoted as M1-M5. To simplify the analysis, the influence of the coupling inductance leakage in the embodiment of this invention is ignored in the steady-state analysis, and the transformer is regarded as an ideal transformer. At the same time, the losses in power devices and circuits are not considered. Since the time of M1, M3, and M4 within a steady-state operating cycle is very short, the focus is on analyzing the two main operating modes, M2 and M5, where the switching transistor is on (M2) and off (M5). The specific analysis is as follows:

[0062] M2 [t1, t2]: such as Figure 4 As shown, the turn-on signal arrives before time t1, and the voltage on semiconductor switch PST1 begins to drop to 0 before time t1. At time t1, the voltage drops to 0, and its current is affected by C1 and L... k The resonance begins, and the sine wave gradually rises. At this time, the anode of D1 is pulled low, and D1 is cut off due to the reverse voltage drop. The coupling inductor begins to store energy, and the energy on CI3 is transferred to capacitor C2 through the switching transistor D2 for energy storage. The output diode D... o It is cut off by the reverse voltage drop. When the current on C2 rises in a sin waveform, it begins to decrease linearly until it drops to 0.

[0063] M5 [t4, t5]: such as Figure 5 As shown, at this time, the semiconductor switch PST1 is in the off state. Since capacitors C1 and C2 are respectively connected to the leakage inductance L on the coupled inductors CI2 and CI3, 2k L 3k Maintaining resonance, the first output filter capacitor C o1 Leakage inductance L on coupled inductor CI1 1k Maintaining resonance, such that diodes D1 and D2... o Within this short time, the current rapidly drops to 0, achieving ZVZCS turn-off; diode D2 is subjected to capacitance C. o2 Voltage V co2 Reverse bias; regulated input power supply V g Together with capacitor C1, energy is transferred to energy storage inductor L1 and magnetizing inductor L2. m Magnetizing inductance L m The energy is then transferred to the coupling inductors CI2 and CI3, as well as the load, thus increasing the magnetizing inductor current i. Lm It continues to decrease until it reaches 0.

[0064] In this embodiment of the invention, the input filter inductor L under the working modes of M2 and M5 is... in The voltage changes across the three-winding coupled inductors CI1, CI2, and CI3 are analyzed using the volt-second balance rule to derive the output voltage V of the converter. o And the expression for the output gain B:

[0065]

[0066]

[0067] in, The voltage of the DC regulated input power supply is n1 = R2 / R1, n2 = R3 / R1, R1, R2, and R3 are the number of turns of the first coupling inductor CI1, the second coupling inductor CI2, and the third coupling inductor CI3, respectively, and D is the duty cycle of the semiconductor switch PST1.

[0068] Based on output voltage gain analysis, this invention uses the turns ratios n1 and n2 of the three-winding coupled inductor and the duty cycle D of the switching transistor as gain adjustment parameters to enhance the converter's boost capability and broaden the adjustable range of the output voltage. This converter can achieve a high boost ratio while keeping the duty cycle D below 0.5, thereby avoiding extreme duty cycle conditions and improving system reliability. In one specific embodiment, when n1=0.5, n2=3, and the duty cycle D is 0.4, the converter's output voltage gain B can reach approximately 11.7 times.

[0069] In this embodiment of the invention, the voltage stress of the switching transistor and the diode is calculated based on the output gain B, as shown in the following expression:

[0070]

[0071] in, V is the voltage across semiconductor switch PST1. D1 V D2 V Do These are the first passive switch D1, the second passive switch D2, and the output diode D. o Voltage at both ends.

[0072] By analyzing capacitors C1 and C2 under the operating modes of M2 and M5, and the filter unit C... o1 C o2 By applying the ampere-second balance rule to the current change, the current stress of the converter's switching transistor and diode is obtained, as expressed below:

[0073]

[0074] in, I is the current flowing through the semiconductor switch PST1. D1 I D2 I Do The current flows through the first passive switch D1, the second passive switch D2, and the output diode D, respectively. o The current, This is the output current flowing through the load.

[0075] In the converter system of this invention, based on voltage and current stress analysis, the input-output boost gain B = 11.7, and the coupling inductor turns ratios n1 = 0.5 and n2 = 3 are set for performance verification experiments. Under these conditions, the voltage stress curves of each active device under different duty cycles D are obtained (based on the output voltage V). o (Based on) such as Figure 6 As shown. By Figure 6 It can be seen that under these operating conditions, the voltage stress borne by each power device remains at a low level. This voltage stress characteristic helps to reduce circuit conduction losses, improve the safety and reliability of circuit operation, and suppress electromagnetic interference generated during converter operation.

[0076] With a duty cycle D = 0.5, the relationship between the voltage gain of the converter of this invention and the transformer turns ratio and duty cycle is as follows: Figure 7 As shown in the figure, the vertical axis G represents the voltage gain.

[0077] In this embodiment of the invention, MATLAB is used, and the output load is set to 1444 Ω, meaning the converter operates normally at 100W power. At 0.2s, the output load jumps to 722 Ω, meaning the converter's output power jumps to 200W, and then jumps back to 1444 Ω at 0.4s. Figure 8 It can be seen that at 0.2s, when the output load becomes heavier, the output voltage jumps to approximately 355V and recovers to the set voltage of 380V after about 15ms; at 0.4s, when the output load returns to a light load, the output voltage jumps to approximately 405V and recovers to the set voltage of 380V after about 15ms. The feasibility of the closed-loop method using the AI ​​control operator, obtained by simulating load change disturbances, is verified.

[0078] Based on the above theoretical analysis and simulation results, it can be confirmed that the converter provided by this invention possesses high voltage gain capability and flexible boost regulation characteristics, while maintaining high conversion efficiency throughout the entire operating range. All key performance indicators meet the design goals, achieving the invention's objectives of high gain, high efficiency, and high reliability.

[0079] Compared with existing technologies, this invention employs a structure combining a common-core multi-winding coupled inductor and a voltage multiplier, which effectively improves voltage gain and reduces device stress, thereby enhancing the converter's performance and control flexibility. Specifically, by introducing a three-winding coupled inductor, the single duty cycle D control of the voltage gain is extended to be jointly adjusted by three control factors: the duty cycle D and the coupled inductor turns ratios n1 and n2. This significantly enhances the freedom and range of the converter's boost regulation, effectively avoiding the problem of the switching transistor having a limited duty cycle when achieving high voltage gain. Simultaneously, the voltage multiplier unit further enhances the system's boost capability, enabling it to stably output a higher voltage even under lower duty cycle and coupled inductor turns ratio conditions, thus balancing system efficiency and overall performance improvement.

[0080] This invention significantly improves the overall performance of the converter through structural innovation and parameter optimization. Specifically, by optimizing the traditional Boost structure, it achieves the dual advantages of low input current ripple and low voltage stress on switching devices, effectively improving system stability and device reliability. Furthermore, by rationally constructing the resonant circuit parameters between the Boost unit and the coupled inductor, without adding additional components, the switching device current operates in a quasi-resonant state, enabling the switching devices to achieve zero-voltage start-up during turn-on or zero-current turn-off during turn-off, thereby significantly reducing switching losses and significantly improving the overall system efficiency.

[0081] This invention also introduces a deep learning-based adaptive control mechanism. Through AI control operators and a deep learning controller, PID control parameters are tuned in real time, dynamically adjusting the switching strategy of the semiconductor switching transistors, significantly improving the system's dynamic response and steady-state accuracy. Compared to conventional PID control, this invention increases the control degrees of freedom from 3 to 6. K can be calculated using the control algorithm by only referencing the desired input voltage. d0 K i0 K p0 Then, the correction amount ΔK is obtained through deep learning technology. d0 ΔK i0 ΔK p0 The control method of this invention avoids the problem of obtaining PID parameters through trial and error in engineering, thus fully optimizing the control strategy. This not only improves control accuracy but also enhances system robustness and anti-interference capabilities.

[0082] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0083] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0084] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0085] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0086] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. An AI self-tuning DC-DC converter system based on a coupled voltage doubler resonant network, characterized in that, It includes input filtering and switching circuits, charging and discharging circuits, voltage multiplier filter output circuits, and a deep learning control subsystem; The input filtering and switching circuit includes a DC regulated input power supply, an input filter inductor, and a semiconductor switching transistor connected in series. The charging and discharging circuit includes a first energy storage capacitor, a second coupling inductor, a stray inductor, and a first coupling inductor connected in series. The starting end of the first energy storage capacitor is connected to a semiconductor switch. Magnetizing inductors are connected in parallel on both sides of the first coupling inductor. The anode of the first passive switch is connected to the starting end of the first input energy storage capacitor, and the cathode of the first passive switch is connected to the same-name end of the first coupling inductor. The voltage multiplier filter output circuit includes a third coupling inductor, a second energy storage capacitor, a second passive switch, an output diode, a first output filter capacitor, and a second output filter capacitor. The same-name terminal of the third coupling inductor is connected to the same-name terminal of the first coupling inductor, and the opposite-name terminal of the third coupling inductor is connected to one end of the second energy storage capacitor. The other end of the second energy storage capacitor is connected to the cathode of the second passive switch and the anode of the output diode, respectively. The cathode of the output diode is connected to one end of the second output filter capacitor. The same-name terminal of the third coupling inductor, the anode of the second passive switch, and the other end of the second output filter capacitor are respectively connected to one end of the first output filter capacitor. The other end of the first output filter capacitor is connected to the negative terminal of the DC regulated input power supply. The input terminal of the deep learning control subsystem is connected to a resistor, which is connected in parallel across the first and second output filter capacitors. The output terminal of the deep learning control subsystem is connected to a semiconductor switching transistor. The deep learning control subsystem includes an output voltage sampling module, a reference voltage setting module, an error comparator, an integrator, a differentiator, an AI control operator module, a deep learning controller, and a modulator. The output voltage sampling module acquires the resistor voltage, the reference voltage setting module sets the desired voltage, the error comparator compares the resistor voltage and the desired voltage to obtain a voltage error signal, and the AI ​​control operator module outputs a first reference control parameter K based on the voltage error signal. d0 Second reference control parameter K i0 Third reference control parameter K p0 The deep learning controller is used to control the first baseline parameter K. d0 Second reference control parameter K i0 Third reference control parameter K p0 Output the first correction amount ΔK d0 Second correction amount ΔK i0 The third correction amount ΔK p0 The integrator and differentiator are used to integrate and differentiate the voltage error signal, respectively, and the modulator is used to control the voltage error signal according to the first reference parameter K. d0 Second reference control parameter K i0 Third reference control parameter K p0 First correction amount ΔK d0 Second correction amount ΔK i0 The third correction amount ΔK p0 The voltage error signal after differentiation and the voltage error signal after integration are output as PWM signals to control the semiconductor switching transistors.

2. The AI ​​self-tuning DC-DC converter system according to claim 1, characterized in that, The first reference control parameter K d0 Second reference control parameter K i0 Third reference control parameter K p0 With the first correction amount ΔK d0 Second correction amount ΔK i0 The third correction amount ΔK p0 By successively subtracting, the corrected first control parameter K is obtained. d Second control parameter K i The third control parameter K p The voltage error signal is integrated using an integrator and then compared with the second control parameter K. i Multiplying them together yields the second intermediate quantity K. i The voltage error signal is differentiated using a differentiator and then compared with the first control parameter K. d Multiplying them together yields the first intermediate quantity K. d ', the third control parameter K p The third intermediate quantity K is obtained by multiplying it by the voltage error signal. p ', will K d '、K i '、K p Input a PWM modulator, and use the PWM modulator to output a PWM signal for controlling the semiconductor switching transistor.

3. The AI ​​self-tuning DC-DC converter system according to claim 1, characterized in that, The first, second, and third coupled inductors are three-winding coupled inductors with a common magnetic core. The number of turns of the coils of the first, second, and third coupled inductors are R1, R2, and R3, respectively. n1 = R2 / R1 and n2 = R3 / R1 are defined, where n1 and n2 are real numbers not equal to 1.

4. The AI ​​self-tuning DC-DC converter system according to claim 1, characterized in that, The semiconductor switching transistor is an insulated gate bipolar transistor or a metal-oxide-semiconductor field-effect transistor.

5. The AI ​​self-tuning DC-DC converter system according to claim 1, characterized in that, The semiconductor switch also includes a parasitic diode, with the anode of the parasitic diode connected to the source of the semiconductor switch and the cathode of the parasitic diode connected to the gate of the semiconductor switch. The drain of the semiconductor switch is connected to the output of the deep learning control subsystem, and the switching on and off of the parasitic diode is controlled by receiving the PWM signal output by the deep learning control subsystem.

6. The AI ​​self-tuning DC-DC converter system according to claim 1, characterized in that, The voltage gain of the system is controlled by coordinating the turns ratio of the first, second, and third coupled inductors and the duty cycle of the semiconductor switch.

7. The AI ​​self-tuning DC-DC converter system according to claim 6, characterized in that, The DC-DC converter system output voltage V o The expression for the output gain B is as follows: ； ； in, The voltage of the DC regulated input power supply is given by n1 and n2, which are real numbers not equal to 1. n1 = R2 / R1, n2 = R3 / R1, R1, R2 and R3 are the number of turns of the first coupling inductor CI1, the second coupling inductor CI2 and the third coupling inductor CI3, respectively, and D is the duty cycle of the semiconductor switch PST1.

8. The AI ​​self-tuning DC-DC converter system according to claim 1, characterized in that, By adjusting the parameters of the magnetizing inductor, stray inductor, first energy storage capacitor, buffer capacitor of semiconductor switch, and first output filter capacitor, the semiconductor switch is turned on under zero voltage conditions under the action of the drive signal; and through the resonance effect between the first coupling inductor, second coupling inductor, third coupling inductor and each capacitor, the first passive switch, second active switch and output switch are turned off with zero voltage and zero current when the semiconductor switch is turned on.

9. The AI ​​self-tuning DC-DC converter system according to claim 1, characterized in that, In a DC-DC converter system, a first energy storage capacitor, a first coupling inductor, and a first passive switch constitute a first clamping unit; a third coupling inductor, a second passive switch, and a second energy storage capacitor constitute a second clamping unit; and a third coupling inductor, a second energy storage capacitor, an output diode, and a second output filter capacitor constitute a third clamping unit. The three clamping units interact in steady state. The first clamping unit stores energy in the first energy storage capacitor and controls the voltage stress across the semiconductor switch. The second clamping unit and the third clamping unit absorb the leakage inductance energy of the third coupling inductor and limit the voltage stress on the second passive switch and the output diode.

Images