Neural network-based five-phase permanent magnet synchronous motor control method for electric vehicle

Abstract

The invention discloses a neural network-based five-phase permanent magnet synchronous motor control method for an electric vehicle, and belongs to the technical field of motor control. The method comprises the following steps: acquiring current operation parameters of the five-phase permanent magnet synchronous motor, and acquiring actual dq-axis current values of a primary plane and a cubic plane through a coordinate transformation module; adjusting a given torque signal by using a limiting module; inputting the adjusted torque signal and the flux linkage limit into a neural network module, and outputting dq-axis reference current of a primary plane and a cubic plane; through the current control module and the weak magnetic feedback module, the following of a current reference value and the adjustment of a flux linkage limit are realized; and finally, a pulse width modulation signal is generated through the SVPWM module, and the five-phase silicon carbide MOSFET driving module is driven to control the motor. According to the method, through feature extraction and interactive calculation of the neural network, torque optimization, current harmonic suppression and smooth switching of MTPA, field weakening control and MTPV modes in a full-speed domain range are realized, and the efficiency and operation stability of the motor are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of motor control, and in particular to a five-phase permanent magnet synchronous motor control method for electric vehicles based on a neural network. Background Art

[0002] With the rapid development of the new energy electric vehicle industry, higher requirements are placed on the efficient and stable control of drive motors. Five-phase permanent magnet synchronous motors have gradually become an important choice in electric vehicle drive systems due to their high power density, lower torque pulsation and good fault tolerance. Compared with traditional three-phase motors, five-phase permanent magnet synchronous motors not only improve output power and efficiency by increasing the number of phases, but also significantly reduce performance degradation under fault conditions, making system operation more reliable. However, the control technology of five-phase permanent magnet synchronous motors in the field of electric vehicle applications still faces many challenges and shortcomings.

[0003] First, the current components of the five-phase permanent magnet synchronous motor include the primary and tertiary dq-axis components, and its multi-degree-of-freedom nonlinear control characteristics increase the complexity of the control algorithm. Although traditional strategies based on proportional integral (PI) control and vector control have been widely used in three-phase motors, they show limitations in complex dynamic control scenarios of five-phase motors. For example, when the load changes or weak magnetic extended operation with high dynamic response requirements are in place, it is difficult for traditional control algorithms to achieve precise regulation of the current of each axis, resulting in large fluctuations in the output torque and poor operating stability.

[0004] Secondly, there is a significant third harmonic current component in the five-phase permanent magnet synchronous motor, which may cause an increase in high-order harmonic losses and total harmonic distortion in actual operation, and have an adverse effect on motor efficiency and running stability. Although existing technologies attempt to reduce the impact of harmonics through current optimization distribution strategies, their optimization models are mostly static or linear approximations, and fail to fully consider the dynamic harmonic characteristics under complex working conditions, and therefore cannot achieve the organic unity of harmonic suppression and efficiency optimization within the full speed range of electric vehicles.

[0005] In addition, in the control of electric vehicle motor drive systems, in order to achieve full-speed operation, the maximum torque current ratio (MTPA) mode, weak magnetic mode and maximum torque voltage ratio (MTPV) mode can be used. Under different operating conditions, these three modes need to be dynamically switched according to the operating state to achieve optimal utilization of energy efficiency. However, existing control methods generally adopt preset mode switching rules or linear interpolation strategies, lack of in-depth modeling of nonlinear operating characteristics and dynamic electromagnetic coupling effects, resulting in an unsmooth switching process and even control fluctuations, affecting the operating stability of the motor.

[0006] In addition, in response to the complex requirements of high-dimensional control of five-phase permanent magnet synchronous motors, existing control methods generally improve control accuracy by introducing multivariable optimization algorithms or large-scale computing models. However, these methods often have high computational complexity, poor real-time performance, and are difficult to adapt to the on-board controller environment of resource-constrained electric vehicles. This limits their promotion and application in actual industrial scenarios.

[0007] In recent years, artificial intelligence and machine learning technologies have gradually demonstrated superior performance in complex system control. Neural networks have good adaptability and generalization capabilities when dealing with high-dimensional nonlinear systems, can dynamically adjust control parameters and learn complex characteristics, and lay a good foundation for solving the above problems. However, in the field of existing five-phase permanent magnet synchronous motor control, the research and application of intelligent algorithms are still in the initial stage. Existing work is mostly focused on the optimization of a single working condition, and a set of efficient full-speed intelligent control solutions has not yet been formed. Summary of the invention

[0008] In view of the shortcomings of the existing technology in the above-mentioned background technology, the present invention proposes a five-phase permanent magnet synchronous motor control method for electric vehicles based on a neural network, which realizes the precise tracking of the primary and tertiary plane dq-axis current operating points based on the neural network, and realizes smooth switching and precise control of the three operating modes of MTPA, weak magnetic control, and MTPV in the full speed range.

[0009] The technical solution to achieve the purpose of the present invention is:

[0010] The control method of a five-phase permanent magnet synchronous motor for electric vehicles based on a neural network has the following steps:

[0011] Step S1, obtaining current parameters of the five-phase permanent magnet synchronous motor, the current parameters including the five-phase current i a 、i b 、i c 、i d 、i e , five-phase voltage v a 、v b 、v c 、v d 、v e And the speed ω e ; The five-phase current i a 、i b 、i c 、i d 、i e And five-phase voltage v a 、v b 、v c 、v d 、v e Input five-phase coordinate transformation module;

[0012] Step S2: Input given torque T * And the flux limit λ obtained in step S6 lim To the limiting module, the limiting module is T * Processing is performed and a processed given torque signal is output;

[0013] Step S3: The given torque signal processed in step S2 and the actual primary plane dq axis current value i obtained in step S4 are d1 、i q1 and the actual three-dimensional plane dq axis current value i d3 、i q3 And the flux limit λ obtained in step S6 lim Input to the neural network module, output a plane d-axis reference current i d1 * , primary plane q-axis reference current i q1 * , three-dimensional plane d-axis reference current i d3 * , three-dimensional plane q-axis reference current i q3 * ;

[0014] Step S4: The five-phase coordinate transformation module transforms the input five-phase current i a 、i b 、i c 、i d 、i e Transformed into the actual primary plane dq axis current value i d1 、i q1 and the actual three-dimensional plane dq axis current value i d3 、i q3 ;

[0015] Step S5: The primary plane d-axis reference current i obtained in step S3 is d1 * , primary plane q-axis reference current i q1 * , three-dimensional plane d-axis reference current i d3 * , three-dimensional plane q-axis reference current i q3 * And the actual primary plane dq axis current value i obtained in step S4 d1 、i q1 and the actual three-dimensional plane dq axis current value i d3 、i q3 Input current control module; the current control module converts i d1 * 、i q1 * 、id3 * 、i q3 * with i d1 、i q1 、i d3 、i q3 Compare, generate control signal through control algorithm, adjust actual current to track reference current; current control module adopts PI control, calculate v d1 、v q1 、v d3 、v q3 , input weak magnetic feedback module and five-phase coordinate inverse transformation module; the PI control formula is: The K p is the proportionality coefficient, the K i is the integration coefficient;

[0017] Step S6: Obtain the DC bus voltage V dc , V dc , speed ω e , the current control module calculates v d1 、v q1 、v d3 、v q3 Input weak magnetic feedback module, weak magnetic feedback module output flux limit λ lim ;

[0018] Step S7, the five-phase coordinate inverse transformation module converts the v calculated in step S5 into d1 、v q1 、v d3 、v q3 Converted into a plane αβ axis voltage v α1 、v β1 and the cubic plane αβ axis voltage v α3 、v β3 , input to SVPWM module;

[0019] Step S8: The SVPWM module generates a voltage v according to the primary plane αβ axis voltage v α1 、v β1 and the cubic plane αβ axis voltage v α3 、v β3 Generate pulse width modulation signal and input it to the five-phase silicon carbide MOSFET driver module;

[0020] Step S9, the power tube driving module in the five-phase silicon carbide MOSFET driving module receives the pulse width modulation signal, and controls the on and off of the SiC MOSFET in the five-phase full-bridge control module according to the pulse width modulation signal, thereby realizing the control of the five-phase winding current of the motor and driving the permanent magnet synchronous motor to operate; the SiC MOSFET is a silicon carbide metal oxide semiconductor field effect transistor.

[0021] Furthermore, the limiting module in step S2 is used to limit the given torque T * , ensuring that the given torque T provided to the neural network module * The maximum allowable torque command is lower than the flux limit; if the desired torque exceeds the limit range, it is maintained at the maximum allowable torque under the flux limit;

[0022] The five-phase coordinate transformation module in step S1 includes a coordinate transformation matrix T C / P , used to implement Clark transform and Park transform;

[0023] The five-phase coordinate inverse transformation module in step S5 includes a coordinate transformation matrix T 反P , used to implement the inverse Park transform.

[0024] Furthermore, the neural network module in step S3 includes an input preprocessing module, a feature extraction module, a feature interaction module, an adaptive reference current adjustment module, and an output module; the input feature of the neural network module is a given torque T * , flux limit λ lim , actual primary plane dq axis current value i d1 、i q1 and the actual three-dimensional plane dq axis current value i d3 、i q3 , the output characteristic is the primary plane d-axis reference current i d1 * , primary plane q-axis reference current i q1 * , three-dimensional plane d-axis reference current i d3 * , three-dimensional plane q-axis reference current i q3 * .

[0025] Furthermore, the input preprocessing module normalizes the input features and synthesizes the plane input features and cubic plane input features Said is the normalized given torque, is the normalized flux linkage limit, is the normalized actual primary plane d-axis current value, is the normalized actual primary plane q-axis current value, is the normalized actual cubic plane d-axis current value, is the normalized actual cubic plane q-axis current value.

[0026] Furthermore, the feature extraction module is divided into a primary plane feature extraction module and a tertiary plane feature extraction module;

[0027] The structure of the primary plane feature extraction module is as follows:

[0028] First layer feature extraction:

[0029] Said is the weight matrix of the first layer of the primary plane, is the bias vector of the first layer of the primary plane; is the output feature of the first layer of the plane;

[0030] Second layer feature extraction:

[0031] Said is the weight matrix of the second layer of the primary plane, is the bias vector of the second layer of the primary plane; is the output feature of the second layer of the first plane;

[0032] The third layer feature extraction:

[0033] Said is the weight matrix of the third layer of the first plane, is the bias vector of the third layer of the primary plane; Output features for a single plane;

[0034] The structure of the three-dimensional plane feature extraction module is as follows:

[0035] First layer feature extraction:

[0036] Said is the weight matrix of the first layer of the cubic plane, is the bias vector of the first layer of the cubic plane; is the output feature of the first layer of the cubic plane;

[0037] Second layer feature extraction:

[0038] Said is the weight matrix of the second layer of the cubic plane, is the bias vector of the second layer of the cubic plane; is the output feature of the second layer of the cubic plane;

[0039] The third layer feature extraction:

[0040] Said is the weight matrix of the third layer of the cubic plane, is the bias vector of the third layer of the cubic plane; Output features for cubic planes.

[0041] Furthermore, the feature interaction module is used to fuse the output features of a plane and cubic plane output features To establish a coupling relationship between the two, the construction steps are as follows:

[0042] Step S3-1-1: Construct fusion feature vector H fus ;

[0043] Step S3-1-2: Construct the first layer of interactive calculation:

[0044] Step S3-1-3: Construct the second layer of interactive calculation:

[0045] Step S3-1-4: construct output interaction features:

[0046] Said is the first layer interaction weight matrix, is the first layer interaction bias vector, is the first layer interaction feature, is the second layer interaction weight matrix, is the second layer interaction bias vector, is the second-layer interaction feature, is the third-layer interaction weight matrix, is the third layer interaction bias vector, H int is the output interaction feature.

[0047] Furthermore, the adaptive reference current adjustment module is based on the output interaction characteristic H int and Dynamically generate a normalized reference current. The steps are as follows:

[0048] Step S3-2-1, generate dynamic adjustment coefficient:

[0049] The W adis the dynamic weight matrix, b ad is the dynamic bias vector, is the dynamic adjustment coefficient of the primary plane d axis, is the dynamic adjustment coefficient of the primary plane q axis, is the dynamic adjustment coefficient of the d-axis of the cubic plane, is the dynamic adjustment coefficient of the q-axis in the cubic plane; The range of is [0, 1];

[0050] Step S3-2-2, output normalized reference current:

[0051]

[0052] The output module denormalizes the normalized reference current and outputs a plane d-axis reference current i d1 * , primary plane q-axis reference current i q1 * , three-dimensional plane d-axis reference current i d3 * , three-dimensional plane q-axis reference current i q3 * .

[0053] Furthermore, the neural network module in step S3 outputs a plane d-axis reference current i d1 * , primary plane q-axis reference current i q1 * , three-dimensional plane d-axis reference current i d3 * , three-dimensional plane q-axis reference current i q3 * , to achieve accurate tracking of the dq-axis current operating points in the primary and tertiary planes, and to achieve smooth switching and precise control of the three operating modes of MTPA, weak magnetic control, and MTPV in the full speed range; the primary plane d-axis reference current i d1 * , primary plane q-axis reference current i q1 * Used to optimize the torque output and efficiency of electric vehicles; the three-dimensional plane d-axis reference current i d3 * , three-dimensional plane q-axis reference current i q3 * The optimal allocation strategy is used to suppress the influence of subharmonics and reduce the total harmonic distortion of current.

[0054] Furthermore, the five-phase silicon carbide MOSFET drive module in step S8 includes a five-phase full-bridge control module and a power tube drive module; the five-phase full-bridge control module is a five-phase full-bridge inverter circuit of a permanent magnet synchronous motor; the power tube drive module includes ten SiC MOSFET dual-source complementary coupled gate drive circuits; the SiC MOSFET dual-source complementary coupled gate drive circuit includes a coupled inductor primary side and a coupled inductor secondary side, wherein the coupled inductor primary side includes: power supply V1, winding L1, switch tube Q1; the coupled inductor secondary side includes: switch tube S+, switch tube S-, auxiliary switch tube Q sub , winding L2, diode D, gate resistor R g , voltage source V+ and voltage source V-; the switch tube Q1, switch tube S+, switch tube S-, auxiliary switch tube Q sub It is a MOSFET switch tube;

[0055] The five-phase full-bridge control module includes first to tenth power tubes, wherein one end of the first power tube, the third power tube, the fifth power tube, the seventh power tube, and the ninth power tube are respectively connected to the five-phase windings A, B, C, D, and E, and the other ends are connected to the positive electrode of the DC bus; one end of the second power tube, the fourth power tube, the sixth power tube, the eighth power tube, and the tenth power tube are respectively connected to the five-phase windings A, B, C, D, and E, and the other ends are connected to the negative electrode of the DC bus; the first to tenth power tubes are SiC MOSFETs;

[0056] In the power tube driving module, on the primary side of the coupled inductor, the positive electrode of the power supply V1 is connected to the upper end of the winding L1, the lower end of the winding L1 is connected to the drain of the switch tube Q1, and the source of the switch tube Q1 is connected to the negative electrode of the power supply V1; on the secondary side of the coupled inductor, the switch tube Q1 is connected to the negative electrode of the power supply V1. sub The source is only connected to the anode of the diode D, the cathode of the diode D is only connected to the upper end of the winding L2, the lower end of the winding L2 is respectively connected to the negative electrode of the voltage source V+, the voltage source V- and the source of the SiC MOSFET, and the gate of the SiC MOSFET is only connected to the gate resistor R g Connected, the connection point is R g Right side, R g On the left side, there is the drain of the switch tube S-, the source of the switch tube S+, and the switch tube Q sub The drain of the switch tube S- is connected to the positive electrode of the voltage source V-, and the drain of the switch tube S+ is connected to the positive electrode of the voltage source V+; the same-name ends of the winding L1 and the winding L2 are both located at the upper end.

[0057] Furthermore, in the five-phase silicon carbide MOSFET drive module, the power tube drive module is responsible for controlling the on and off of the ten power tubes in the five-phase full-bridge control module, and the five-phase full-bridge control module is responsible for controlling the five-phase windings of the permanent magnet synchronous motor;

[0058] A complete opening and closing process includes the following steps:

[0059] Step S9-1: First, close S+ and connect the gate of SiC MOSFET to the gate resistor R g Connected to the voltage source V+, so that the gate voltage of SiCMOSFET V gs Maintain the voltage value V at the voltage source V+ on , at this time, the gate capacitance is in a charging state. When the charging is completed, it reaches a steady state, and no additional current flows into or out of the gate. That is, the gate current i of the SiC MOSFET g =0, SiC MOSFET is turned on, duration Δt last When SiCMOSFET is turned on, the switch tube Q1 on the primary side of the coupled inductor is closed. At this time, the current flows from the power supply V1 into the winding L1 on the primary side of the coupled inductor, and the auxiliary switch tube Q sub In the disconnected state, although the secondary winding L2 generates electromotive force due to electromagnetic induction, no actual current path is formed;

[0060] Step S9-2, duration Δt last After the end, close the auxiliary switch tube Q sub At this time, Q1 remains on and there is a short overlap time Δt between the two. overlop ; At the same time, disconnect S+ and cut off the voltage source V+ power supply;

[0061] Step S9-3: Overlap time Δt overlop After the end, disconnect Q1, but switch tube Q sub Still closed, as Q1 is turned off, the current in the primary winding is interrupted, and the magnetic flux accumulated in the coupled inductor core generates an induced current in the secondary winding, which discharges the gate-source capacitance of the SiC MOSFET, turning it off;

[0062] Step S9-4: After the SiC MOSFET is completely turned off, disconnect Q sub , close S-, the voltage source V- makes the gate-source voltage of SiC MOSFET stable at the voltage value V of the voltage source V- off , to ensure that the SiC MOSFET is reliably turned off.

[0063] Compared with the prior art, the five-phase permanent magnet synchronous motor control method for electric vehicles based on neural network according to the present invention has the following beneficial effects:

[0064] 1. The neural network module proposed in this patent realizes accurate tracking of the dq-axis current operating points in the primary and cubic planes based on the neural network, effectively overcoming the problem that the traditional control algorithm is difficult to achieve accurate adjustment of the current of each axis when applied to the five-phase permanent magnet synchronous motor. Compared with the traditional control algorithm, the output torque pulsation is small and the operation stability is high.

[0065] 2. This patent effectively utilizes the good adaptability and generalization ability of neural networks in processing high-dimensional nonlinear systems, as well as the ability to dynamically adjust control parameters and learn complex characteristics, to achieve the organic unity of harmonic suppression and efficiency optimization within the full speed range of electric vehicles.

[0066] 3. The neural network module proposed in this patent, through the feature extraction module and the feature interaction module, effectively copes with the complex nonlinear problems in the control of the five-phase permanent magnet synchronous motor, improves the response performance of the motor drive system under nonlinear and dynamically changing conditions, and is based on the adaptive reference current regulation module. It dynamically optimizes the reference current of the primary and cubic plane dq axes according to the input characteristics, significantly improves the current control accuracy, and optimizes the energy efficiency performance of the motor, and realizes smooth switching and precise control of the three operating modes of MTPA, weak magnetic control, and MTPV within the full speed range; the method proposed in this patent is more stable than the traditional method based on preset mode switching rules or linear interpolation.

[0067] 4. Based on the application of neural networks, the present invention adopts SiC MOSFET to replace traditional IGBT as the inverter switch tube to further achieve the organic unity of harmonic suppression and efficiency optimization in the full speed range of electric vehicles, thereby reducing switching losses and improving efficiency to meet the driving performance requirements of the five-phase permanent magnet synchronous motor for electric vehicles.

[0068] 5. The five-phase silicon carbide MOSFET driver module proposed in this patent has excellent high temperature resistance and can work stably at higher temperatures. When the electric vehicle is running at high power or the heat dissipation conditions are relatively poor, it can still maintain reliable performance, reduce the risk of failure due to overheating, and improve the stability and reliability of the system.

[0069] 6. The SiC MOSFET dual-source complementary coupled gate drive circuit proposed in this patent uses a coupled inductor to effectively control the voltage and current change rate of the power tube during the shutdown process, achieves accurate gate charge extraction, reduces switching losses, and improves switching performance; by reasonably adjusting the timing of the drive circuit and introducing controllable delays, the voltage balance is further optimized, reducing the impact of factors such as device parameter mismatch and parasitic capacitance, so that the drive circuit can still maintain good performance when facing complex working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 This is a flow chart of a five-phase permanent magnet synchronous motor control method for electric vehicles based on a neural network proposed by the present invention;

[0071] Figure 2 This is a control block diagram of a five-phase permanent magnet synchronous motor for electric vehicles based on a neural network proposed by the present invention;

[0072] Figure 3 This is a schematic diagram of the five-phase silicon carbide MOSFET driving module proposed by the present invention;

[0073] Figure 4 This is a topology diagram of the SiC MOSFET dual-source complementary coupled gate drive circuit proposed by the present invention;

[0074] Figure 5 A current flow path diagram of the SiC MOSFET dual-source complementary coupled gate drive circuit from time t1 to time t2 proposed by the present invention;

[0075] Figure 6 A current flow path diagram of the SiC MOSFET dual-source complementary coupled gate drive circuit proposed by the present invention from time t2 to time t3;

[0076] Figure 7 FIG. 4 is a current flow path diagram of the SiC MOSFET dual-source complementary coupled gate drive circuit proposed in the present invention from time t3 to time t4. DETAILED DESCRIPTION

[0077] The following is in conjunction with the instruction manual Figure 1-Figure 7 , the present invention is described based on embodiments, but the present invention is not limited to these embodiments. In the detailed description of the present invention below, some specific details are described in detail. For those skilled in the art, the present invention can be fully understood without the description of these details.

[0078] like Figure 1 , Figure 2 As shown in the figure, the control method of the five-phase permanent magnet synchronous motor for electric vehicles based on neural network is as follows:

[0079] Step S1, obtaining current parameters of the five-phase permanent magnet synchronous motor, the current parameters including the five-phase current i a 、i b 、i c 、i d 、i e , five-phase voltage v a 、v b 、v c 、v d 、v e And the speed ωe ; The five-phase current i a 、i b 、i c 、i d 、i e And the five-phase voltage v a 、v b 、v c 、v d 、v e Input five-phase coordinate transformation module;

[0080] Step S2: Input given torque T * And the flux limit λ obtained in step S6 lim To the limiting module, the limiting module is T * Processing is performed and a processed given torque signal is output;

[0081] Step S3: The given torque signal processed in step S2 and the actual primary plane dq axis current value i obtained in step S4 are d1 、i q1 and the actual three-dimensional plane dq axis current value i d3 、i q3 And the flux limit λ obtained in step S6 lim Input to the neural network module, output a plane d-axis reference current i d1 * , primary plane q-axis reference current i q1 * , three-dimensional plane d-axis reference current i d3 * , three-dimensional plane q-axis reference current i q3 * ;

[0082] Step S4: The five-phase coordinate transformation module transforms the input five-phase current i a 、i b 、i c 、i d 、i e Transformed into the actual primary plane dq axis current value i d1 、i q1 and the actual three-dimensional plane dq axis current value i d3 、i q3 ;

[0083] Step S5: The primary plane d-axis reference current i obtained in step S3 is d1 * , primary plane q-axis reference current i q1 * , three-dimensional plane d-axis reference current i d3 *, three-dimensional plane q-axis reference current i q3 * And the actual primary plane dq axis current value i obtained in step S4 d1 、i q1 and the actual three-dimensional plane dq axis current value i d3 、i q3 Input current control module; the current control module converts i d1 * 、i q1 * 、i d3 * 、i q3 * with i d1 、i q1 、i d3 、i q3 Compare, generate control signal through control algorithm, adjust actual current to track reference current; current control module adopts PI control, calculate v d1 、v q1 、v d3 、v q3 , input weak magnetic feedback module and five-phase coordinate inverse transformation module; the PI control formula is: The K p is the proportionality coefficient, the K i is the integration coefficient;

[0085] Step S6: Obtain the DC bus voltage V dc , V dc , speed ω e , the current control module calculates v d1 、v q1 、v d3 、v q3 Input weak magnetic feedback module, weak magnetic feedback module output flux limit λ lim ;

[0086] Step S7, the five-phase coordinate inverse transformation module converts the v calculated in step S5 into d1 、v q1 、v d3 、v q3 Converted into primary plane αβ axis voltage v α1 、v β1 and the cubic plane αβ axis voltage v α3 、v β3 , input to SVPWM module;

[0087] Step S8: The SVPWM module generates a voltage v according to the primary plane αβ axis voltage v α1 、vβ1 and the cubic plane αβ axis voltage v α3 、v β3 Generate pulse width modulation signal and input it to the five-phase silicon carbide MOSFET driver module;

[0088] Step S9, the power tube driving module in the five-phase silicon carbide MOSFET driving module receives the pulse width modulation signal, and controls the on and off of the SiC MOSFET in the five-phase full-bridge control module according to the pulse width modulation signal, thereby realizing the control of the five-phase winding current of the motor and driving the permanent magnet synchronous motor to operate; the SiC MOSFET is a silicon carbide metal oxide semiconductor field effect transistor.

[0089] Furthermore, the limiting module in step S2 is used to limit the given torque T * , ensuring that the given torque T provided to the neural network module * The maximum allowable torque command is lower than the flux limit; if the desired torque exceeds the limit range, it is maintained at the maximum allowable torque under the flux limit;

[0090] The five-phase coordinate transformation module in step S1 includes a coordinate transformation matrix T C / P , used to implement Clark transform and Park transform;

[0091] The five-phase coordinate inverse transformation module in step S5 includes a coordinate transformation matrix T 反P , used to implement the inverse Park transform.

[0092] Furthermore, the neural network module in step S3 includes an input preprocessing module, a feature extraction module, a feature interaction module, an adaptive reference current adjustment module, and an output module; the input feature of the neural network module is a given torque T * , flux limit λ lim , actual primary plane dq axis current value i d1 、i q1 and the actual three-dimensional plane dq axis current value i d3 、i q3 , the output characteristic is the primary plane d-axis reference current i d1 * , primary plane q-axis reference current i q1 * , three-dimensional plane d-axis reference current i d3 * , three-dimensional plane q-axis reference current i q3 * .

[0093] Furthermore, the input preprocessing module normalizes the input features and synthesizes the plane input features and cubic plane input features Said is the normalized given torque, is the normalized flux linkage limit, is the normalized actual primary plane d-axis current value, is the normalized actual primary plane q-axis current value, is the normalized actual cubic plane d-axis current value, is the normalized actual cubic plane q-axis current value.

[0094] Furthermore, the feature extraction module is divided into a primary plane feature extraction module and a tertiary plane feature extraction module;

[0095] The structure of the primary plane feature extraction module is as follows:

[0096] First layer feature extraction:

[0097] Said is the weight matrix of the first layer of the primary plane, is the bias vector of the first layer of the primary plane; is the output feature of the first layer of the plane;

[0098] Second layer feature extraction:

[0099] Said is the weight matrix of the second layer of the primary plane, is the bias vector of the second layer of the primary plane; is the output feature of the second layer of the first plane;

[0100] The third layer feature extraction:

[0101] Said is the weight matrix of the third layer of the first plane, is the bias vector of the third layer of the primary plane; Output features for a single plane;

[0102] The structure of the three-dimensional plane feature extraction module is as follows:

[0103] First layer feature extraction:

[0104] Said is the weight matrix of the first layer of the cubic plane, is the bias vector of the first layer of the cubic plane; is the output feature of the first layer of the cubic plane;

[0105] Second layer feature extraction:

[0106] Said is the weight matrix of the second layer of the cubic plane, is the bias vector of the second layer of the cubic plane; is the output feature of the second layer of the cubic plane;

[0107] The third layer feature extraction:

[0108] Said is the weight matrix of the third layer of the cubic plane, is the bias vector of the third layer of the cubic plane; Output features for cubic planes.

[0109] Furthermore, the feature interaction module is used to fuse the output features of a plane and cubic plane output features To establish a coupling relationship between the two, the construction steps are as follows:

[0110] Step S3-1-1: Construct fusion feature vector H fus ;

[0111] Step S3-1-2: Construct the first layer of interactive calculation:

[0112] Step S3-1-3: Construct the second layer of interactive calculation:

[0113] Step S3-1-4: construct output interaction features:

[0114] Said is the first layer interaction weight matrix, is the first layer interaction bias vector, is the first layer interaction feature, is the second layer interaction weight matrix, is the second layer interaction bias vector, is the second-layer interaction feature, is the third-layer interaction weight matrix, is the third layer interaction bias vector, H int is the output interaction feature.

[0115] Furthermore, the adaptive reference current adjustment module is based on the output interaction characteristic H int and Dynamically generate a normalized reference current. The steps are as follows:

[0116] Step S3-2-1, generate dynamic adjustment coefficient:

[0117] The W ad is the dynamic weight matrix, b ad is the dynamic bias vector, is the dynamic adjustment coefficient of the primary plane d axis, is the dynamic adjustment coefficient of the primary plane q axis, is the dynamic adjustment coefficient of the d-axis of the cubic plane, is the dynamic adjustment coefficient of the q-axis in the cubic plane; The range of is [0, 1];

[0118] Step S3-2-2, output normalized reference current:

[0119]

[0120] The output module denormalizes the normalized reference current and outputs a plane d-axis reference current i d1 * , primary plane q-axis reference current i q1 * , three-dimensional plane d-axis reference current i d3 * , three-dimensional plane q-axis reference current i q3 * .

[0121] Furthermore, the neural network module in step S3 outputs a plane d-axis reference current i d1 * , primary plane q-axis reference current i q1 * , three-dimensional plane d-axis reference current i d3 * , three-dimensional plane q-axis reference current i q3 * , to achieve accurate tracking of the dq-axis current operating points in the primary and tertiary planes, and to achieve smooth switching and precise control of the three operating modes of MTPA, weak magnetic control, and MTPV in the full speed range; the primary plane d-axis reference current i d1 * , primary plane q-axis reference current i q1 * Used to optimize the torque output and efficiency of electric vehicles; the three-dimensional plane d-axis reference current i d3 * , three-dimensional plane q-axis reference current i q3 * The optimal allocation strategy is used to suppress the influence of subharmonics and reduce the total harmonic distortion of current.

[0122] Further, such as Figure 3 As shown, the five-phase silicon carbide MOSFET driving module in step S8 includes a five-phase full-bridge control module and a power tube driving module; the five-phase full-bridge control module is a five-phase full-bridge inverter circuit of a permanent magnet synchronous motor; the power tube driving module includes ten SiCMOSFET dual-source complementary coupling gate driving circuits; Figure 4 As shown, the SiCMOSFET dual-source complementary coupled gate drive circuit includes a coupled inductor primary side and a coupled inductor secondary side, wherein the coupled inductor primary side includes: power supply V1, winding L1, switch tube Q1; the coupled inductor secondary side includes: switch tube S+, switch tube S-, auxiliary switch tube Q sub , winding L2, diode D, gate resistor R g , voltage source V+ and voltage source V-; the switch tube Q1, switch tube S+, switch tube S-, auxiliary switch tube Q sub It is a MOSFET switch tube;

[0123] The five-phase full-bridge control module includes first to tenth power tubes, wherein one end of the first power tube, the third power tube, the fifth power tube, the seventh power tube, and the ninth power tube are respectively connected to the five-phase windings A, B, C, D, and E, and the other ends are connected to the positive electrode of the DC bus; one end of the second power tube, the fourth power tube, the sixth power tube, the eighth power tube, and the tenth power tube are respectively connected to the five-phase windings A, B, C, D, and E, and the other ends are connected to the negative electrode of the DC bus; the first to tenth power tubes are SiC MOSFETs;

[0124] In the power tube driving module, on the primary side of the coupled inductor, the positive electrode of the power supply V1 is connected to the upper end of the winding L1, the lower end of the winding L1 is connected to the drain of the switch tube Q1, and the source of the switch tube Q1 is connected to the negative electrode of the power supply V1; on the secondary side of the coupled inductor, the switch tube Q1 is connected to the negative electrode of the power supply V1. sub The source is only connected to the anode of the diode D, the cathode of the diode D is only connected to the upper end of the winding L2, the lower end of the winding L2 is respectively connected to the negative electrode of the voltage source V+, the voltage source V- and the source of the SiC MOSFET, and the gate of the SiC MOSFET is only connected to the gate resistor R g Connected, the connection point is R g Right side, R g On the left side, there is the drain of the switch tube S-, the source of the switch tube S+, and the switch tube Q sub The drain of the switch tube S- is connected to the positive electrode of the voltage source V-, and the drain of the switch tube S+ is connected to the positive electrode of the voltage source V+; the same-name ends of the winding L1 and the winding L2 are both located at the upper end.

[0125] Furthermore, in the five-phase silicon carbide MOSFET drive module, the power tube drive module is responsible for controlling the on and off of the ten power tubes in the five-phase full-bridge control module, and the five-phase full-bridge control module is responsible for controlling the five-phase windings of the permanent magnet synchronous motor;

[0126] A complete opening and closing process includes the following steps:

[0127] Step S9-1: First, close S+ and connect the gate of SiC MOSFET to the gate resistor R g Connected to the voltage source V+, so that the gate voltage of SiCMOSFET V gs Maintain the voltage value V at the voltage source V+ on , at this time, the gate capacitance is in a charging state. When the charging is completed, it reaches a steady state, and no additional current flows into or out of the gate. That is, the gate current i of the SiC MOSFET g =0, SiC MOSFET is turned on, duration Δt last When SiCMOSFET is turned on, the switch tube Q1 on the primary side of the coupled inductor is closed. At this time, the current flows from the power supply V1 into the winding L1 on the primary side of the coupled inductor, and the auxiliary switch tube Q sub In the disconnected state, although the secondary winding L2 generates electromotive force due to electromagnetic induction, no actual current path is formed;

[0128] Step S9-2, duration Δt last After the end, close the auxiliary switch tube Q sub At this time, Q1 remains on and there is a short overlap time Δt between the two. overlop ; At the same time, disconnect S+ and cut off the voltage source V+ power supply;

[0129] Step S9-3: Overlap time Δt overlop After the end, disconnect Q1, but switch tube Q sub Still closed, as Q1 is turned off, the current in the primary winding is interrupted, and the magnetic flux accumulated in the coupled inductor core generates an induced current in the secondary winding, which discharges the gate-source capacitance of the SiC MOSFET, turning it off;

[0130] Step S9-4: After the SiC MOSFET is completely turned off, disconnect Q sub , close S-, the voltage source V- makes the gate-source voltage of SiC MOSFET stable at the voltage value V of the voltage source V- off , to ensure that the SiC MOSFET is reliably turned off.

[0131] The time of implementing step S9-1 is t1, the time of implementing step S9-2 is t2, the time of implementing step S9-3 is t3, and the time of implementing step S9-4 is t4; the current flow path at the time t1-t2 is as follows: Figure 5 As shown, the current flow path at time t2-t3 is as follows Figure 6 As shown, the current flow path at time t3-t4 is as follows Figure 7 shown.

[0132] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A five-phase permanent magnet synchronous motor control method for electric vehicles based on a neural network, characterized in that: The method steps are as follows: Step S1, obtaining current parameters of the five-phase permanent magnet synchronous motor, the current parameters including the five-phase current i a 、i b 、i c 、i d 、i e , five-phase voltage v a 、v b 、v c 、v d 、v e And the speed ω e ; The five-phase current i a 、i b 、i c 、i d 、i e And the five-phase voltage v a 、v b 、v c 、v d 、v e Input five-phase coordinate transformation module; Step S2: Input given torque T * And the flux limit λ obtained in step S6 lim To the limiting module, the limiting module is T * Processing is performed and a processed given torque signal is output; Step S3: The given torque signal processed in step S2 and the actual primary plane dq axis current value i obtained in step S4 are d1 、i q1 and the actual three-dimensional plane dq axis current value i d3 、i q3 And the flux limit λ obtained in step S6 lim Input to the neural network module, output a plane d-axis reference current i d1 * , primary plane q-axis reference current i q1 * , three-dimensional plane d-axis reference current i d3 * , three-dimensional plane q-axis reference current i q3 * ; Step S4: The five-phase coordinate transformation module transforms the input five-phase current i a 、i b 、i c 、i d 、i e Transformed into the actual primary plane dq axis current value i d1 、i q1 and the actual three-dimensional plane dq axis current value i d3 、i q3 ; Step S5: The primary plane d-axis reference current i obtained in step S3 is d1 * , primary plane q-axis reference current i q1 * , three-dimensional plane d-axis reference current i d3 * , three-dimensional plane q-axis reference current i q3 * And the actual primary plane dq axis current value i obtained in step S4 d1 、i q1 and the actual three-dimensional plane dq axis current value i d3 、i q3 Input current control module; The current control module will i d1 * 、i q1 * 、i d3 * 、i q3 * with i d1 、i q1 、i d3 、i q3 Compare, generate control signal through control algorithm, adjust actual current to track reference current; current control module adopts PI control, calculate v d1 、v q1 、v d3 、v q3 , input weak magnetic feedback module and five-phase coordinate inverse transformation module; the PI control formula is: The K p is the proportionality coefficient, the K i is the integration coefficient; Step S6: Obtain the DC bus voltage V dc , V dc , speed ω e , the current control module calculates v d1 、v q1 、v d3 、v q3 Input weak magnetic feedback module, weak magnetic feedback module output flux limit λ lim ; Step S7, the five-phase coordinate inverse transformation module converts the v calculated in step S5 into d1 、v q1 、v d3 、v q3 Converted into a plane αβ axis voltage v α1 、v β1 and the cubic plane αβ axis voltage v α3 、v β3 , input to SVPWM module; Step S8: The SVPWM module generates a voltage v according to the primary plane αβ axis voltage v α1 、v β1 and the cubic plane αβ axis voltage v α3 、v β3 Generate pulse width modulation signal and input it to the five-phase silicon carbide MOSFET driver module; Step S9, the power tube driving module in the five-phase silicon carbide MOSFET driving module receives the pulse width modulation signal, and controls the on and off of the SiCMOSFET in the five-phase full-bridge control module according to the pulse width modulation signal, thereby realizing the control of the five-phase winding current of the motor and driving the permanent magnet synchronous motor to operate; the SiC MOSFET is a silicon carbide metal oxide semiconductor field effect transistor.

2. The neural network-based five-phase permanent magnet synchronous motor control method for electric vehicles according to claim 1, characterized in that: The limiting module in step S2 is used to limit the given torque T * , ensuring that the given torque T provided to the neural network module * The maximum allowable torque command is lower than the flux limit; if the desired torque exceeds the limit range, it is maintained at the maximum allowable torque under the flux limit; The five-phase coordinate transformation module in step S1 includes a coordinate transformation matrix T C / P , used to implement Clark transform and Park transform; The five-phase coordinate inverse transformation module in step S5 includes a coordinate transformation matrix T 反P , used to implement the inverse Park transform.

3. The neural network-based five-phase permanent magnet synchronous motor control method for electric vehicles according to claim 1, characterized in that: The neural network module in step S3 includes an input preprocessing module, a feature extraction module, a feature interaction module, an adaptive reference current adjustment module, and an output module; the input feature of the neural network module is a given torque T * , flux limit λ lim , actual primary plane dq axis current value i d1 、i q1 and the actual three-dimensional plane dq axis current value i d3 、i q3 , the output characteristic is the primary plane d-axis reference current i d1 * , primary plane q-axis reference current i q1 * , three-dimensional plane d-axis reference current i d3 * , three-dimensional plane q-axis reference current i q3 * .

4. The neural network-based five-phase permanent magnet synchronous motor control method for electric vehicles according to claim 3, characterized in that: The input preprocessing module normalizes the input features and synthesizes the plane input features and cubic plane input features Said is the normalized given torque, is the normalized flux linkage limit, is the normalized actual primary plane d-axis current value, is the normalized actual primary plane q-axis current value, is the normalized actual cubic plane d-axis current value, is the normalized actual cubic plane q-axis current value.

5. The neural network-based five-phase permanent magnet synchronous motor control method for electric vehicles according to claim 3, characterized in that: The feature extraction module is divided into a primary plane feature extraction module and a tertiary plane feature extraction module; The structure of the primary plane feature extraction module is as follows: First layer feature extraction: Said is the weight matrix of the first layer of the primary plane, is the bias vector of the first layer of the primary plane; is the output feature of the first layer of the plane; Second layer feature extraction: Said is the weight matrix of the second layer of the primary plane, is the bias vector of the second layer of the primary plane; is the output feature of the second layer of the first plane; The third layer feature extraction: Said is the weight matrix of the third layer of the first plane, is the bias vector of the third layer of the primary plane; Output features for a single plane; The structure of the three-dimensional plane feature extraction module is as follows: First layer feature extraction: Said is the weight matrix of the first layer of the cubic plane, is the bias vector of the first layer of the cubic plane; is the output feature of the first layer of the cubic plane; Second layer feature extraction: Said is the weight matrix of the second layer of the cubic plane, is the bias vector of the second layer of the cubic plane; is the output feature of the second layer of the cubic plane; The third layer feature extraction: Said is the weight matrix of the third layer of the cubic plane, is the bias vector of the third layer of the cubic plane; Output features for cubic planes.

6. The neural network-based five-phase permanent magnet synchronous motor control method for electric vehicles according to claim 3, characterized in that: The feature interaction module is used to fuse the primary plane output features and cubic plane output features To establish a coupling relationship between the two, the construction steps are as follows: Step S3-1-1: Construct fusion feature vector H fus ; Step S3-1-2: Construct the first layer of interactive calculation: Step S3-1-3: Construct the second layer of interactive calculation: Step S3-1-4: construct output interaction features: W1 int is the first layer interaction weight matrix, is the first layer interaction bias vector, is the first layer interaction feature, is the second layer interaction weight matrix, is the second layer interaction bias vector, is the second-layer interactive feature, W3 int is the third-layer interaction weight matrix, is the third layer interaction bias vector, H int is the output interaction feature.

7. The neural network-based five-phase permanent magnet synchronous motor control method for electric vehicles according to claim 3, characterized in that: The adaptive reference current regulation module is based on the output interaction characteristic H int and To dynamically generate a normalized reference current, follow these steps: Step S3-2-1, generate dynamic adjustment coefficient: The W ad is the dynamic weight matrix, b ad is the dynamic bias vector, is the dynamic adjustment coefficient of the primary plane d axis, is the dynamic adjustment coefficient of the q-axis of the primary plane, is the dynamic adjustment coefficient of the d-axis of the cubic plane, is the dynamic adjustment coefficient of the q-axis in the cubic plane; The range of is [0, 1]; Step S3-2-2, output normalized reference current: The output module denormalizes the normalized reference current and outputs a plane d-axis reference current i d1 * , primary plane q-axis reference current i q1 * , three-dimensional plane d-axis reference current i d3 * , three-dimensional plane q-axis reference current i q3 * .

8. The neural network-based five-phase permanent magnet synchronous motor control method for electric vehicles according to claim 1, characterized in that: The neural network module in step S3 outputs a plane d-axis reference current i d1 * , primary plane q-axis reference current i q1 * , three-dimensional plane d-axis reference current i d3 * , three-dimensional plane q-axis reference current i q3 * , to achieve accurate tracking of the dq-axis current operating points in the primary and tertiary planes, and to achieve smooth switching and precise control of the three operating modes of MTPA, weak magnetic control, and MTPV in the full speed range; the primary plane d-axis reference current i d1 * , primary plane q-axis reference current i q1 * Used to optimize the torque output and efficiency of electric vehicles; the three-dimensional plane d-axis reference current i d3 * , three-dimensional plane q-axis reference current i q3 * The optimal allocation strategy is used to suppress the influence of subharmonics and reduce the total harmonic distortion of current.

9. The neural network-based five-phase permanent magnet synchronous motor control method for electric vehicles according to claim 1, characterized in that: The five-phase silicon carbide MOSFET driving module in step S8 includes a five-phase full-bridge control module and a power tube driving module; the five-phase full-bridge control module is a five-phase full-bridge inverter circuit of a permanent magnet synchronous motor; the power tube driving module includes ten SiCMOSFET dual-source complementary coupling gate driving circuits; The SiC MOSFET dual-source complementary coupled gate drive circuit includes a coupled inductor primary side and a coupled inductor secondary side, wherein the coupled inductor primary side includes: a power supply V1, a winding L1, and a switch tube Q1; the coupled inductor secondary side includes: a switch tube S+, a switch tube S-, an auxiliary switch tube Q sub , winding L2, diode D, gate resistor R g , voltage source V+ and voltage source V-; the switch tube Q1, switch tube S+, switch tube S-, auxiliary switch tube Q sub It is a MOSFET switch tube; The five-phase full-bridge control module includes first to tenth power tubes, wherein one end of the first power tube, the third power tube, the fifth power tube, the seventh power tube, and the ninth power tube are respectively connected to the five-phase windings A, B, C, D, and E, and the other ends are connected to the positive electrode of the DC bus; one end of the second power tube, the fourth power tube, the sixth power tube, the eighth power tube, and the tenth power tube are respectively connected to the five-phase windings A, B, C, D, and E, and the other ends are connected to the negative electrode of the DC bus; the first to tenth power tubes are SiC MOSFETs; In the power tube driving module, on the primary side of the coupled inductor, the positive electrode of the power supply V1 is connected to the upper end of the winding L1, the lower end of the winding L1 is connected to the drain of the switch tube Q1, and the source of the switch tube Q1 is connected to the negative electrode of the power supply V1; on the secondary side of the coupled inductor, the switch tube Q1 is connected to the negative electrode of the power supply V1. sub The source is only connected to the anode of the diode D, the cathode of the diode D is only connected to the upper end of the winding L2, the lower end of the winding L2 is respectively connected to the negative electrode of the voltage source V+, the voltage source V- and the source of the SiC MOSFET, and the gate of the SiC MOSFET is only connected to the gate resistor R g Connected, the connection point is R g Right side, R g On the left side, there is the drain of the switch tube S-, the source of the switch tube S+, and the switch tube Q sub The drain of the switch tube S- is connected to the positive electrode of the voltage source V-, and the drain of the switch tube S+ is connected to the positive electrode of the voltage source V+; the same-name ends of the winding L1 and the winding L2 are both located at the upper end.

10. The neural network-based five-phase permanent magnet synchronous motor control method for electric vehicles according to claim 1, characterized in that: In the five-phase silicon carbide MOSFET drive module, the power tube drive module is responsible for controlling the on and off of the ten power tubes in the five-phase full-bridge control module, and the five-phase full-bridge control module is responsible for controlling the five-phase windings of the permanent magnet synchronous motor; A complete opening and closing process includes the following steps: Step S9-1: First, close S+ and connect the gate of SiC MOSFET to the gate resistor R g Connected to the voltage source V+, so that the gate voltage of SiCMOSFET V gs Maintain the voltage value V of the voltage source V+ on , at this time, the gate capacitance is in a charging state. When the charging is completed, it reaches a steady state, and no additional current flows into or out of the gate. That is, the gate current i of the SiC MOSFET g =0, SiCMOSFET is turned on, duration Δt last When SiCMOSFET is turned on, the switch Q1 on the primary side of the coupled inductor is closed. At this time, the current flows from the power supply V1 into the winding L1 on the primary side of the coupled inductor, and the auxiliary switch Q1 on the secondary side is turned on. sub In the disconnected state, although the secondary winding L2 generates electromotive force due to electromagnetic induction, no actual current path is formed; Step S9-2, duration Δt last After the end, close the auxiliary switch tube Q sub At this time, Q1 remains on and there is a short overlap time Δt between the two. overlop ; At the same time, disconnect S+ and cut off the voltage source V+ power supply; Step S9-3: Overlap time Δt overlop After the end, disconnect Q1, but switch tube Q sub Still closed, as Q1 is turned off, the current in the primary winding is interrupted, and the magnetic flux accumulated in the coupled inductor core generates an induced current in the secondary winding, which discharges the gate-source capacitance of the SiC MOSFET, turning it off; Step S9-4: After the SiC MOSFET is completely turned off, disconnect Q sub , close S-, the voltage source V- makes the gate-source voltage of SiC MOSFET stable at the voltage value V of the voltage source V- off , to ensure that the SiC MOSFET is reliably turned off.

Images