Cascaded multilevel converter control strategy based on recurrent neural network

Through the combination of intelligent control strategies based on recursive neural networks and cascading converters, the oscillation and stability problems of traditional controllers under complex voltage conversion requirements are solved, and efficient voltage adjustment and power transmission under wide voltage range are achieved.

CN118944406APending Publication Date: 2024-11-12HOHAI UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411087340.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-08
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

Traditional cascading converter controllers are prone to oscillation and unstable when facing complex voltage conversion requirements, and parameter design depends on the accuracy of system modeling, making it difficult to adapt to system parameter changes.

Method used

Adopting an intelligent control strategy based on recursive neural network, the duty cycle of the switch tube is dynamically adjusted through the recursive neural network controller, combined with a non-isolated FSBB converter, an isolated CLLC resonant converter and a three-phase PWM converter to achieve wide input voltage Voltage adjustment and power flow in both directions in range.

Benefits of technology

It improves the robustness and dynamic and static performance of the system, and realizes efficient voltage adjustment and power transmission under different motor operating conditions and battery energy storage.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118944406A_ABST
    Figure CN118944406A_ABST
Patent Text Reader

Abstract

The invention provides a cascaded multilevel converter control strategy based on a recurrent neural network, and aims to provide a new control mode. The system comprises a front-stage four-switch Buck-Boost (FSBB) converter, an intermediate-stage symmetrical CLLC resonant converter, a rear-stage three-phase PWM converter and a permanent magnet synchronous motor. A recurrent neural network controller is trained by adopting a Levenberg-Marquardt (LM) optimization method, a weight matrix obtained offline is used as a neural network initialization weight, and a traditional PI controller is replaced by combining an error back propagation method of a neural network to realize control of the multi-stage converter. The beneficial effects of the invention are that the technical scheme provided by the invention can improve the robust performance and dynamic response capability of the system, and solves the problems of parameter coupling and nonlinearity in the control of the multistage converter.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of intelligent control of cascade converters, in particular to a control method of cascade multi-level converters based on recursive neural networks. Background Art

[0002] The continuous progress of power electronics technology has made cascade converters widely used in many fields. This converter can effectively convert DC power of different voltage levels, so it is favored in various fields. It has significant advantages such as high efficiency and high power density, so it is widely used in wireless power transmission, distributed power generation systems, DC uninterruptible power supply systems, smart grids, and electric vehicle power drives.

[0003] As the voltage range increases or the difference between input and output voltages increases, single-level power converters are often unable to effectively perform the required voltage conversion tasks, so cascade converters become essential. Currently common cascade converters usually include two key levels: non-isolated converters and isolated converters. Non-isolated converters are mainly responsible for voltage conversion tasks, while isolated converters assume the important responsibility of voltage isolation to ensure the safety of the system under a wide range of input voltage conditions. This cascade structure performs well in dealing with complex voltage conversion requirements.

[0004] Once the parameters of traditional linear controllers such as PI controllers are determined, they cannot be changed. When the system conditions or parameters change, the system may oscillate or even become unstable, which in turn affects the operating state of the motor. In addition, the design process of controller parameters is relatively complex and depends on the accuracy of system modeling. Cascade converters have the characteristics of multivariable and nonlinear, while neural networks have the ability of self-adaptation, self-learning and generalization, and can approximate any nonlinear function. These characteristics make them widely used in the field of nonlinear control. Compared with traditional control methods, neural networks do not rely on the mathematical model of the system. Using neural networks to control cascade converter systems can greatly improve the robustness and dynamic and static performance of the system. Summary of the invention

[0005] In order to solve the problems of shortcomings in traditional control, the present invention provides a cascade converter control strategy based on a recursive neural network to achieve adjustment of the output bus voltage under a wide input voltage range, thereby realizing motor drive and battery energy storage.

[0006] In order to achieve the above purpose, the technical solution of the present invention is as follows:

[0007] The system topology includes a non-isolated adjustable voltage FSBB converter, an isolated CLLC resonant converter for wireless power transmission, a three-phase PWM converter and a permanent magnet motor PMSM;

[0008] The FSBB converter is used as the front-stage voltage regulation circuit, the CLLC resonant converter is used as the middle-stage isolation circuit, and the three-phase PWM converter is used as the rear-stage load control circuit. The three-stage system can realize the forward and reverse bidirectional flow of power.

[0009] The front-stage FSBB converter consists of an inductor L1 and four switch tubes S1, S2, S3, and S4. The switch tubes S1 and S2 form a group of bridge arms, which are Buck units, and the switch tubes S3 and S4 form a group of bridge arms, which are Boost units. The boost and buck functions of the FSBB circuit are realized through the cooperation of the two groups of switch tubes:

[0010] The intermediate CLLC resonant converter consists of switch tube S5□S12, loosely coupled high-frequency rotary transformer T1, primary resonant capacitor C r1 , primary resonant inductor L r1 , secondary side resonant capacitor C r2 And the secondary resonant capacitor L r2 Composition: Switch tube S5□S8 forms the primary side, loosely coupled high-frequency rotary transformer T1, primary resonant capacitor C r1 , primary resonant inductor L r1 , secondary side resonant capacitor C r2 And the secondary resonant capacitor L r2 The resonant cavity is formed, L2 is the transformer excitation inductance, the switch tube S9□S12 forms the secondary side, and the loosely coupled high-frequency rotary transformer T1 realizes the wireless power transmission function;

[0011] The subsequent load converter is composed of a switch tube S13□S18 and a PMSM, wherein the switch tube S13□S18 forms a three-phase PWM converter, which can be used for rectification or inversion.

[0012] When the power flows in the forward direction, the drive system can dynamically adjust the bus voltage U according to the operating conditions of the permanent magnet motor. dc ;

[0013] According to the operating conditions of the permanent magnet motor, the bus 3 voltage Udc can be dynamically adjusted by switching the working mode of FSBB (Buck mode, Boost mode), and the voltage E and current iL of bus 2 and the actual voltage Udc of bus 3 are sent to the intelligent controller 1. The duty cycle of the corresponding switch tube of FSBB is dynamically adjusted through the intelligent controller 1, and the bus 2 voltage E is adjusted, and then the bus 3 voltage Udc is adjusted to achieve maximum efficiency transmission.

[0014] The primary full bridge H1 of the intermediate-stage CLLC converter constitutes the inverter side, and the four switching tubes of the secondary full bridge H2 are all turned off, and the diodes of the switching tubes constitute the rectifier side.

[0015] When the power flows forward, the three-phase PWM converter at the rear stage adopts the SVPWM control method, and the upper and lower switch tubes of the same bridge arm are complementary turned on to realize the inverter function to drive the motor. The three-phase current signal of the motor is collected by the current sensor, and the current signal under the dq coordinate axis is obtained through coordinate transformation. The speed signal of the motor is obtained by the position sensor, and the speed signal and the current signal are input into the intelligent controller 2 to obtain the motor drive signal.

[0016] When the power flows in reverse, the permanent magnet motor acts as a generator, and the three-phase PWM converter adopts the SVPWM control method to realize the rectification function and transmit the electric energy to the CLLC converter. The current sensor collects the current signal output by the motor, and the current signal under the dq coordinate axis is obtained through coordinate transformation. The voltage signal of the bus 3 is obtained by the voltage sensor, and the voltage signal and the current signal are input into the intelligent controller 3 to control the output voltage of the rectifier.

[0017] The intermediate-stage CLLC converter full-bridge H2 serves as the inverter side, and the full-bridge H1 serves as the rectifier side. The switch control methods of the inverter bridge H2 and the rectifier bridge H1 are consistent with those when the power flows in the forward direction.

[0018] When the power flows in the reverse direction, the drive system can adjust the bus 1 voltage U0 according to the battery pack power situation. According to the battery pack power situation, the bus 1 voltage U0 is dynamically adjusted by switching the working mode of FSBB (Buck mode, Boost mode), and the bus 2 voltage E and current iL and bus 1 voltage U0 are input into the intelligent controller 4. By dynamically adjusting the duty cycle of the corresponding switch tube of FSBB, the bus 1 voltage U0 is adjusted to realize charging of the battery pack.

[0019] The resonant cavity part in the intermediate stage CLLC resonant converter, the resonant inductor L r1 and resonant capacitor C r1 Can be connected in parallel, resonant inductor L r2 and resonant capacitor C r2 Can be combined in parallel to allow bidirectional power flow.

[0020] The intelligent controller of this system is composed of a recursive neural network. The weight matrix of the neural network is obtained through offline training using the LM algorithm. Combined with the error back propagation method of the neural network, the weight update matrix is ​​adjusted in the direction of decreasing the evaluation function to adjust the weight matrix in the recursive neural network controller. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is the topology diagram of the cascade converter system.

[0022] Figure 2 This is the control block diagram of the system power forward flow mode.

[0023] Figure 3 This is the control block diagram of the system power reverse flow mode.

[0024] Figure 4 This is the control block diagram of intelligent controller 1.

[0025] Figure 5 This is the control block diagram of intelligent controller 2.

[0026] Figure 6 This is the control block diagram of intelligent controller 3.

[0027] Figure 7 This is the control block diagram of the intelligent controller 4. DETAILED DESCRIPTION

[0028] The specific implementation modes of the present invention are described in detail below in conjunction with the accompanying drawings so that relevant technical personnel can better understand the present invention.

[0029] like Figure 1 The figure shows the general topology of the cascade converter system, including the non-isolated adjustable voltage FSBB converter, the isolated CLLC resonant converter for wireless power transmission, the three-phase PWM converter and the permanent magnet synchronous motor PMSM.

[0030] The FSBB converter is used as the front-stage voltage regulation circuit, the CLLC resonant converter is used as the middle-stage isolation circuit, and the three-phase PWM converter is used as the rear-stage load control circuit. The three-stage system can realize the forward and reverse bidirectional flow of power.

[0031] like Figure 2 As shown in Figure 1, when the system power flows forward, the three-phase PWM converter works in the inverter mode to control the operating condition of the permanent magnet motor and dynamically adjusts the bus voltage U by switching the working mode of FSBB (Buck mode, Boost mode). dc .

[0032] like Figure 4 The figure shows the control block diagram of the intelligent controller 1. The voltage E and current iL of bus 2 and the actual voltage Udc of bus 3 are sent to the intelligent controller 1. The duty cycle of the corresponding switch tube of FSBB is dynamically adjusted by the intelligent controller 1 to adjust the voltage E of bus 2 and then adjust the voltage Udc of bus 3 to achieve maximum efficiency transmission.

[0033] like Figure 5 The figure shows the control block diagram of the intelligent controller 2. The three-phase PWM converter at the rear stage adopts the SVPWM control method. The current sensor collects the current signal of the motor, and the current signal under the dq coordinate axis is obtained through Clark and Park transformation. The speed signal of the motor is obtained by the position sensor, and the speed signal and the q-axis current signal are input into the intelligent controller 2 to obtain the motor drive signal.

[0034] like Figure 3 As shown, when the system power flows in the reverse direction, the drive system can adjust the bus 1 voltage U0 according to the battery pack power status.

[0035] like Figure 6 The control block diagram of the intelligent controller 3 is shown. The permanent magnet motor is used as a generator. The three-phase PWM converter adopts the SVPWM control method to realize the rectification function and transmit the electric energy to the CLLC converter. The current signal output by the motor is collected by the current sensor, and the current signal under the dq coordinate axis is obtained through coordinate transformation. The voltage signal of the bus 3 is obtained by the voltage sensor, and the voltage signal and the current signal are input into the intelligent controller 4 to control the output voltage of the rectifier.

[0036] The CLLC converter full bridge H2 serves as the inverter side, and the full bridge H1 serves as the rectifier side. The switch control method of the inverter bridge H2 and the rectifier bridge H1 is the same as when the power flows in the forward direction.

[0037] like Figure 7 The figure shows the control block diagram of the intelligent controller 4. According to the battery pack power status, the bus 1 voltage U0 is dynamically adjusted by switching the working mode of FSBB (Buck mode, Boost mode), the bus 2 voltage E and current iL and the bus 1 voltage U0 are input into the intelligent controller 3, and the bus 1 voltage U0 is adjusted by dynamically adjusting the duty cycle of the corresponding switch tube of FSBB to realize charging of the battery pack.

[0038] The four intelligent controllers of the present invention are all trained offline through pre-collected system data and optimized through the LM algorithm. When the training performance of the network reaches a satisfactory level, it is deployed in the DSP to control the system.

[0039] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A cascaded multi-level converter control strategy based on a recurrent neural network, characterized in that: Including non-isolated adjustable voltage FSBB converter, isolated CLLC resonant converter for wireless power transmission, three-phase PWM converter and permanent magnet motor PMSM; The FSBB converter is used as the front-stage voltage regulation circuit, the CLLC resonant converter is used as the middle-stage isolation circuit, and the three-phase PWM converter is used as the rear-stage load control circuit. The three-stage system can realize the forward and reverse bidirectional flow of power. The front-stage FSBB converter consists of an inductor L1 and four switch tubes S1, S2, S3, and S4. The switch tubes S1 and S2 form a group of bridge arms, which are Buck units, and the switch tubes S3 and S4 form a group of bridge arms, which are Boost units. The boost and buck functions of the FSBB circuit are realized through the cooperation of the two groups of switch tubes. The intermediate CLLC resonant converter consists of switch tube S5□S12, loosely coupled high-frequency rotary transformer T1, and primary resonant capacitor C r1 , primary resonant inductor L r1 , secondary side resonant capacitor C r2 And the secondary resonant capacitor L r2 Composition: Switch tube S5□S8 forms the primary side, loosely coupled high-frequency rotary transformer T1, primary resonant capacitor C r1 , primary resonant inductor L r1 , secondary side resonant capacitor C r2 And the secondary resonant capacitor L r2 The resonant cavity is formed, L2 is the transformer excitation inductance, the switch tube S9□S12 forms the secondary side, and the loosely coupled high-frequency rotary transformer T1 realizes the wireless power transmission function; The subsequent load converter is composed of a switch tube S13□S18 and a PMSM, wherein the switch tube S13□S18 forms a three-phase PWM converter, which can be used for rectification or inversion.

2. The cascaded multi-level converter control strategy based on recursive neural network according to claim 1, characterized in that: When the power flows in the forward direction, the drive system can dynamically adjust the bus voltage U according to the operating conditions of the permanent magnet motor. dc ; According to the operating conditions of the permanent magnet motor, the bus 3 voltage U can be dynamically adjusted by switching the working mode of FSBB (Buck mode, Boost mode) dc , the voltage E and current i of bus 2 L The actual voltage U of bus 3 dc The voltage is sent to the intelligent controller 1, which dynamically adjusts the duty cycle of the corresponding switch tube of FSBB, adjusts the bus 2 voltage E, and then adjusts the bus 3 voltage U dc , to achieve maximum efficiency transmission; The primary full bridge H1 of the intermediate-stage CLLC converter constitutes the inverter side, and the four switching tubes of the secondary full bridge H2 are all turned off, and the diodes of the switching tubes constitute the rectifier side.

3. The control strategy of cascade converter based on recursive neural network according to claim 1, characterized in that: When the power flows in the forward direction, the three-phase PWM converter at the rear stage adopts the SVPWM control method, and the upper and lower switch tubes of the same bridge arm are turned on complementary to realize the inverter function to drive the motor; The current sensor collects the current signal of the motor, and the current signal under the dq coordinate axis is obtained through coordinate transformation. The position sensor obtains the speed signal of the motor, and the speed signal and the current signal are input into the intelligent controller 2 to obtain the motor drive signal.

4. The cascaded multi-level converter control strategy based on recursive neural network according to claim 1, characterized in that: When the power flows in the reverse direction, the permanent magnet motor acts as a generator, and the three-phase PWM converter adopts the SVPWM control method to realize the rectification function and transmit the electric energy to the CLLC converter; The three-phase current signal output by the motor is collected by the current sensor, and the current signal under the dq coordinate axis is obtained through coordinate transformation. The voltage signal of the bus 3 is obtained by the voltage sensor, and the voltage signal and the current signal are input into the intelligent controller 3 to control the output voltage of the rectifier; The intermediate-stage CLLC converter full-bridge H2 serves as the inverter side, and the full-bridge H1 serves as the rectifier side. The switch control methods of the inverter bridge H2 and the rectifier bridge H1 are consistent with those when the power flows in the forward direction.

5. The cascaded multi-level converter control strategy based on recursive neural network according to claim 1, characterized in that: When the power flows in the reverse direction, the drive system can adjust the bus 1 voltage U0 according to the battery pack power status; According to the battery pack power status, the bus 1 voltage U0 is dynamically adjusted by switching the FSBB working mode (Buck mode, Boost mode), and the bus 2 voltage E and current i L The bus 1 voltage U0 is input into the intelligent controller 4, and the bus 1 voltage U0 is adjusted by dynamically adjusting the duty cycle of the corresponding switch tube of FSBB to charge the battery pack.

6. The control strategy of cascade converter based on recursive neural network according to claim 1, characterized in that: The resonant cavity part in the intermediate stage CLLC resonant converter, the resonant inductor L r1 and resonant capacitor C r1 Can be connected in parallel, resonant inductor L r2 and resonant capacitor C r2 Can be combined in parallel to allow bidirectional power flow.

7. The control strategy of cascade converter based on recursive neural network according to claim 1, characterized in that: The intelligent controller of this system is composed of a recursive neural network. The weight matrix of the neural network is obtained through offline training using the LM algorithm. Combined with the error back propagation method of the neural network, the weight update matrix is ​​adjusted in the direction of decreasing the evaluation function to adjust the weight matrix in the recursive neural network controller.