A method for suppressing torque fluctuations in a hub motor system based on dual-layer control

By leveraging the synergistic effect of the self-disturbance rejection speed loop and the dual-vector model predictive current loop in a two-layer control architecture, the problem of torque fluctuation suppression in hub motors in existing technologies is solved, effectively suppressing speed fluctuations and current harmonics, and improving the stability and reliability of the system.

CN123122345APending Publication Date: 2026-11-17CHONGQING UNIV +2
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611531936.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-30
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies mostly focus on improving a single component, such as the current loop or the speed loop, with little consideration for the coordinated operation of the two components. As a result, it is difficult to effectively suppress torque fluctuations in hub motors under all operating conditions.

Method used

A two-layer control architecture consisting of an upper-layer active disturbance rejection speed loop and a lower-layer dual-vector model predictive current loop is adopted. The upper-layer active disturbance rejection speed loop observes and compensates for the total disturbance in real time through an extended state observer, while the lower-layer dual-vector model predictive current loop selects the optimal voltage vector in each sampling period to suppress current harmonics, and the synergistic effect suppresses torque fluctuations.

Benefits of technology

It significantly reduces speed fluctuations and current harmonics in the hub motor system, improves system stability and reliability, and enhances overall vehicle driving comfort under complex operating conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123122345A_ABST
    Figure CN123122345A_ABST
Patent Text Reader

Abstract

This invention relates to a method for suppressing torque fluctuations in a hub motor system based on dual-layer control, belonging to the field of electric vehicle hub motor drive control technology. The method constructs a dual-layer control architecture consisting of an upper-layer active disturbance rejection speed loop and a lower-layer dual-vector model predictive current loop: the upper layer treats sudden load torque changes as the total system disturbance, observes and feeds forward compensation using an improved extended state observer with proportional feedback of speed observation error, and outputs a quadrature-axis current setpoint after nonlinear error feedback and overcurrent limiting to suppress speed fluctuations caused by load changes; the lower layer selects a voltage vector twice in each sampling cycle, allocates the action time of the two vectors based on the q-axis current deadbeat principle, expands the synthesized voltage vector to any direction and any amplitude, and uses a wave generation module to synthesize dual-vector pulses to drive the three-phase inverter, reducing current harmonics. This invention can effectively reduce torque fluctuations and current harmonics in hub motors, weaken speed fluctuations, and enhance stability and robustness.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of electric vehicle hub motor drive control technology, and relates to a method for suppressing torque fluctuations in a hub motor system based on dual-layer control. Background Technology

[0002] In-wheel motors, due to their compact structure, elimination of the mechanical transmission chain, fast dynamic response, and ease of achieving independent four-wheel drive, have gradually become a research hotspot in electric vehicle drive technology. However, in-wheel motors are directly integrated inside the wheel, directly exposed to road impacts, vibrations, and mud and water erosion, operating in a harsh environment. Simultaneously, electric vehicles experience frequent torque fluctuations during acceleration, braking, hill climbing, and steering, placing higher demands on the control performance of the in-wheel motor system. Because in-wheel motors are direct-drive structures, torque fluctuations in the motor rotor are directly transmitted to the wheels and body without being attenuated by a reduction gear, causing longitudinal vibration and ride discomfort. Long-term exposure to alternating torque also reduces the fatigue life of mechanical components such as wheel bearings and suspension, lowering the reliability of the in-wheel motor system. Therefore, suppressing torque fluctuations in in-wheel motors is crucial for improving the comfort and reliability of in-wheel motor-driven electric vehicles.

[0003] The current loop and speed loop are the two core control components of a permanent magnet synchronous motor speed control system, and their control performance directly determines the final output of the system. Traditional PI control methods are simple in structure and easy to implement in engineering. However, their parameter tuning is highly dependent on engineering experience, making it difficult to simultaneously meet the requirements of dynamic response and steady-state accuracy. Furthermore, they are sensitive to changes in motor parameters and external load disturbances, and are prone to problems such as large overshoot, long settling time, and even oscillation, making them unsuitable for hub motor systems with rapidly changing loads.

[0004] To overcome the shortcomings of traditional PI control, scholars have proposed various nonlinear control methods, such as hysteresis current control, sliding mode control, fuzzy control, and neural network control. Hysteresis current control offers rapid response and strong robustness, but its bang-bang control method and variable switching frequency lead to significant current ripple and electromagnetic noise, increasing inverter switching losses. Sliding mode control is insensitive to parameter changes and external disturbances, but it suffers from inherent chattering issues, with high-frequency chattering directly translating into torque fluctuations. Fuzzy control and neural network control do not rely on precise mathematical models, but they suffer from high computational complexity and poor real-time performance, making them difficult to implement on low-cost control chips.

[0005] Model predictive control (MPC), as a model-based optimization control algorithm, uses a system mathematical model to predict future state changes within a finite time domain. It selects the optimal control quantity by solving the optimization problem online, offering advantages such as fast dynamic response, ease of handling multiple constraints, and direct utilization of the inverter's discrete characteristics. MPC can be divided into continuous control set model predictive control (CCS-MPC) and finite control set model predictive control (FCS-MPC). Finite control set model predictive current control (FCS-MPCC) limits the control set to the seven basic voltage vectors of a two-level inverter, eliminating the need for modulation and offering a simple and intuitive control concept. However, traditional single-vector MPCC only applies one voltage vector per sampling period, and the voltage vector can only be selected in 6 fixed directions, resulting in large steady-state current fluctuations. Although duty cycle MPCC introduces duty cycle adjustment, its second voltage vector is fixed as a zero voltage vector, and the voltage vector can still only adjust its amplitude in 6 fixed directions. Moreover, the duty cycle is calculated after the voltage vector is selected. When the predicted current value after the optimal voltage vector is applied is less than the given value, even if the duty cycle is 1, the current beat-free requirement cannot be achieved, and the global optimality of the selected vector cannot be guaranteed. The current still fluctuates greatly, which in turn causes electromagnetic torque fluctuations.

[0006] In terms of the speed loop, Active Disturbance Rejection Control (ADRC) does not rely on a precise mathematical model of the controlled object. It uses an Extended State Observer (ESO) to observe and compensate for total disturbances inside and outside the system in real time, exhibiting good disturbance rejection performance. However, in a standard ESO, the disturbance observation value is only the result of the observation error integrated over time, resulting in low utilization of the error signal. This leads to a slow observation speed for sudden disturbances, untimely compensation, and short-term fluctuations in speed, which are then transmitted as torque fluctuations through the current loop.

[0007] In summary, existing research mainly focuses on improving a single component, either the current loop or the speed loop, with less consideration given to the synergistic effect between the two components: improving the current loop alone cannot suppress speed fluctuations caused by sudden load changes, and improving the speed loop alone cannot suppress torque pulsations caused by current harmonics. Neither of these approaches can effectively suppress torque fluctuations in hub motors under all operating conditions.

[0008] Therefore, how to provide a control method that can simultaneously suppress the speed fluctuation and current harmonics of the hub motor, thereby reducing torque fluctuation, has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0009] This invention aims to address the problem that existing technologies often focus on improving a single link in the current loop or speed loop, with little consideration for the synergistic cooperation between the two links, making it difficult to effectively suppress torque fluctuations in hub motors under all operating conditions. It provides a torque fluctuation suppression method for hub motor systems based on dual-layer control, which simultaneously suppresses hub motor speed fluctuations and current harmonics, thereby reducing torque fluctuations.

[0010] In view of the above, the technical solution of the present invention is as follows: A torque fluctuation suppression method for a hub motor system based on dual-layer control is proposed, which adopts a dual-layer control architecture consisting of an upper-layer active disturbance rejection speed loop and a lower-layer dual-vector model predictive current loop. The upper-level active disturbance rejection speed loop treats the load torque, viscous friction, and unmodeled dynamics as the total system disturbance and expands them into new state variables. An improved extended state observer with proportional feedback of speed observation error is constructed to observe the motor speed and the total system disturbance in real time. The difference between the speed setpoint and the speed observation value is calculated by the nonlinear fal function error feedback law, and the total disturbance observation value is converted into a compensation current for feedforward compensation. After current limiting, the quadrature axis current setpoint is output. The lower-level dual-vector model predicts the current loop by selecting two optimal voltage vectors in each sampling period based on the quadrature-axis current setpoint, the direct-axis current setpoint, and the quadrature-axis and direct-axis current feedback values, and allocating their action time. The pulse signal is then synthesized by the waveform generation module to drive the three-phase voltage source inverter to control the hub motor.

[0011] Furthermore, a mathematical model of the permanent magnet synchronous motor for hub motor in synchronous rotating dq coordinate system is established, including the stator voltage equation shown in equations (1) and (2), the electromagnetic torque equation shown in equation (3), and the mechanical motion equation shown in equation (4). The stator voltage equation is discretized using the first-order Euler discretization method to obtain the current prediction model shown in equations (5) and (6). The three-phase stator current is collected, and the current components in the two-phase stationary coordinate system are obtained by the Clark transformation shown in equation (7). The d and q axis feedback currents are obtained by combining the rotor electric angle position with the Park transformation shown in equation (8).

[0012] Furthermore, the upper-level active disturbance rejection speed loop classifies the load torque, viscous friction, and unmodeled dynamics as the total system disturbance according to equation (9), and writes the speed loop in the first-order integral cascaded standard form shown in equation (10); constructs an improved extended state observer with proportional feedback of speed observation error as shown in equations (11) to (13), wherein the improved extended state observer expands the total system disturbance into new state variables, and the speed observation value is obtained by the speed observation error through the error feedback gain coefficient. Feedback correction: The total system disturbance observation value is calculated by adjusting the speed observation error via the error feedback gain coefficient. Feedback and proportional feedback gain coefficient Proportional feedback is used for joint correction.

[0013] Furthermore, the characteristic equation of the coefficient matrix of the improved extended state observer is shown in equation (14), and its eigenvalues ​​are uniformly configured in the observation bandwidth. At this point, the observer parameters are tuned according to equation (15); the error feedback gain coefficient... , All are positive real numbers, and the proportional feedback gain coefficient Not greater than the error feedback gain coefficient .

[0014] Furthermore, the difference between the given rotational speed and the observed rotational speed is calculated using the nonlinear fal function error feedback law shown in equation (16). The nonlinear fal function has the characteristics of large error and small gain, and small error and large gain, and its parameters satisfy 0 < <1, The threshold value for the linear segment interval is given by equation (17). After compensation by the total disturbance observation value, the output of the error feedback law is used to obtain the unlimited quadrature axis current given value. Then, after being limited to the maximum allowable quadrature axis current by the overcurrent limiting circuit shown in equation (18), the quadrature axis current given value is output. The direct-axis current setpoint is taken as =0.

[0015] Furthermore, the lower-level dual-vector model predicts the current loop by performing two voltage vector selections in each sampling period. The candidate vectors for the second vector selection are not limited to the zero voltage vector, and the action time of the two voltage vectors is allocated based on the q-axis current deadbeat principle. The value function considers the action time of the voltage vectors while pre-selecting them, and finally the pulse is synthesized by the waveform generation module to drive the three-phase voltage source inverter. A total of 14 current predictions and value function evaluations are performed in each sampling period.

[0016] Furthermore, the first voltage vector selection is specifically as follows: the 6 effective voltage vectors and 1 zero voltage vector of the two-level three-phase voltage source inverter are transformed by Park according to equations (19) and (20) to obtain the corresponding d and q axis voltage components. These components are then substituted into the current prediction model to predict the d and q axis currents at the next sampling time. Finally, the values ​​of the values ​​corresponding to each basic voltage vector are traversed and compared. The voltage vector that minimizes the value function is selected as the first optimal voltage vector. .

[0017] Furthermore, the voltage vector application time allocation is specifically as follows: the first optimal voltage vector... Combined with 7 candidate voltage vectors, the effective time of two voltage vectors in each combination within one sampling period is allocated using the q-axis current deadbeat principle. Let the first optimal voltage vector be... effect Candidate voltage vector action - The q-axis current reaches the given value at the next sampling time, as shown in equation (22); the first optimal voltage vector is obtained by solving equation (23). The duration of action is limited to [0, according to equation (24)]. The current slope under the action of the two voltage vectors is calculated according to equations (25) and (26).

[0018] Furthermore, the second voltage vector selection is as follows: considering the action time of the two voltage vectors, the equivalent d-axis and q-axis voltages are weighted and synthesized according to the action time, as shown in Equation (27). The current prediction model is substituted to predict the current value under the action of the seven vector combinations, and the corresponding value function value is calculated by substituting it into the value function. The candidate vector that minimizes the value function is taken as the second optimal voltage vector. The predicted electromagnetic torque is shown in equation (28); the wave generation module first outputs the first optimal voltage vector within one sampling period. The corresponding three-phase switch state and its duration of action. Then switch to output the second optimal voltage vector. The corresponding three-phase switch status and continue - It generates six switching pulse signals to drive a three-phase voltage source inverter, with the pulse signals of the two switching transistors on the same bridge arm being complementary.

[0019] Furthermore, the torque fluctuation coefficient shown in Equation (29), the current harmonic distortion rate shown in Equation (30), the longitudinal impact degree shown in Equation (31), and the speed fluctuation amount and recovery time shown in Equation (32) are used to evaluate the torque fluctuation suppression effect. Among them, the current harmonic distortion rate is the ratio of the amplitude of all harmonic content to the amplitude of the fundamental wave, the longitudinal impact degree is the rate of change of the longitudinal acceleration of the whole vehicle, the speed fluctuation amount is the maximum deviation of the measured speed from the given speed under the sudden increase and sudden decrease of load torque conditions, and the recovery time is the time required for the speed to re-enter the error band of the given value ±2%.

[0020] Compared with the prior art, the beneficial effects of the present invention are: The upper-layer active disturbance rejection speed loop of this invention treats sudden load torque changes, road surface excitation, viscous friction, and unmodeled dynamics as total disturbances, which are observed and fed forward by the extended state observer in real time. It does not rely on the precise mathematical model of the controlled object and has strong robustness. An improved extended state observer with proportional feedback of speed observation error is introduced. A proportional feedback branch of observation error is added to the disturbance observation channel, and part of the speed feedback gain is contributed to the disturbance observation channel. Without reducing the observation accuracy, the observation pressure of the integral channel is reduced, the observation speed of sudden disturbances is accelerated, and the system can compensate for disturbances more timely. Experiments show that, under the same observation bandwidth, the disturbance observation rise time and fall time of the improved extended state observer are both shorter than those of the standard extended state observer, with a speed increase of about 21% to 29% when the observation bandwidth is small. Compared with traditional PI speed control, the step response overshoot after adopting the self-disturbance rejection speed loop is reduced from about 108 r / min to near zero, and the settling time is shortened from about 126 ms to about 40 ms. When the rated load is suddenly applied, the speed fluctuation is reduced from 26 r / min to 8 r / min, and the recovery time is shortened from 460 ms to 360 ms. The suppression of speed fluctuation directly reduces the fluctuation of electromagnetic torque.

[0021] The lower-level dual-vector model predictive current loop of this invention performs two voltage vector selections in each sampling period. The second voltage vector is not limited to the zero voltage vector, expanding the selection range of the synthesized voltage vector from six fixed directions with adjustable amplitude to any direction and any amplitude. Furthermore, the value function considers the action time of the voltage vector while pre-selecting it, avoiding the problem of the selected vector being no longer optimal due to selecting the vector first and then calculating the duty cycle in duty cycle model predictive current control. This ensures that the final voltage vector combination is globally optimal, resulting in more accurate voltage vector selection. Experiments show that compared with duty cycle model predictive current control, dual-vector model predictive current control can reduce d-axis and q-axis current fluctuations from 0.95A and 1.05A to approximately 0.35A and 0.30A, respectively, and significantly reduce the stator current harmonic distortion rate, thereby effectively suppressing torque fluctuations caused by current harmonics. Moreover, only 14 value function evaluations are required per sampling period, with a computational load far less than the 25 combinations traversed in the generalized dual-vector scheme, making it easy to implement in real-time on a low-cost digital signal processor.

[0022] The upper-layer active disturbance rejection (ADDR) speed loop of this invention suppresses speed fluctuations, while the lower-layer dual-vector model predictive current loop suppresses current harmonics. These two layers work synergistically to suppress torque fluctuations in the hub motor from both the source and propagation path perspectives. Experiments show that, compared to the standard ADDR controller using an improved extended state observer, the ADDR controller employing the improved extended state observer further reduces speed fluctuations during sudden loading from 12 r / min to 8 r / min (a reduction of approximately 33%), and during sudden unloading from 29 r / min to 18 r / min (a reduction of approximately 38%). Recovery times are also shortened from 380 ms and 220 ms to 360 ms and 180 ms, respectively. This dual-layer synergy significantly reduces torque fluctuation coefficients, current harmonic distortion rate, and longitudinal impact while maintaining dynamic response speed, enhancing the stability and robustness of the hub motor system under complex operating conditions, and improving the reliability of the hub motor system and the overall vehicle driving comfort.

[0023] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0024] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 This is an overall flowchart of the torque fluctuation suppression method for a hub motor system based on dual-layer control according to the present invention; Figure 2 This is a block diagram of the dual-layer control method for suppressing torque fluctuations in a hub motor system based on dual-layer control, as described in this invention. Figure 3 This is a structural diagram of the improved extended state observer with proportional feedback of rotational speed observation error introduced in this invention; Figure 4 This is a flowchart of the dual-vector model predictive current loop control of the present invention; Figure 5 A schematic diagram showing the voltage vector selection range of a traditional MPCC, a duty cycle MPCC, and the dual-vector MPCC of this invention; Figure 6 This is a schematic diagram of the timing of dual-vector pulse synthesis according to the present invention; Figure 7 This is a topology diagram of the two-level three-phase voltage source inverter of the present invention; Figure 8 A comparison of the disturbance observation curves between the improved extended state observer and the standard extended state observer; Figure 9This is a comparison chart of the speed response curves for three speed loop control methods. Detailed Implementation

[0025] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0026] Example 1 This embodiment illustrates the complete implementation process of a torque ripple suppression method for a hub motor system based on dual-layer control. Please refer to... Figures 1-7 A method for suppressing torque fluctuations in a hub motor system based on dual-layer control is shown in the flowchart below. Figure 1 As shown, the overall block diagram of the two-layer control is as follows: Figure 2 As shown, it includes the following steps: Step 1: Building the mathematical model of the hub motor and transforming the coordinates of the feedback signal; Step 2: Construction of the upper-layer self-disruption rejection speed loop; Step 3: Building the lower-level dual-vector model to predict the current loop; Step 4: Setting evaluation indicators for torque fluctuation suppression effect.

[0027] This embodiment takes the external rotor permanent magnet synchronous hub motor of a four-wheel independent drive electric vehicle as the control object. The control system is built in the Matlab / Simulink environment and implemented using a digital signal processor.

[0028] like Figure 2 As shown, the upper-layer active disturbance rejection speed loop receives the given speed. Measured speed and quadrature axis feedback current Output quadrature-axis current setpoint Lower-level dual-vector model predicts current loop reception. =0、 And feedback , Electric angular velocity Electric angle position DC bus voltage and sampling period In each sampling period, two optimal voltage vectors are selected and their effective time is assigned. The vector number and first time are output. To the waveform generation module; the waveform generation module synthesizes six-channel switching pulses. ~ The three-phase voltage source inverter drives the hub motor; the motor's three-phase stator current is fed back to the upper and lower level controllers after Clark and Park transformations, forming a closed loop.

[0029] The main parameters of the hub motor are: stator resistance =0.0308Ω, direct-axis inductance =0.316mH, quadrature axis inductance =0.323mH, permanent magnet flux linkage =0.2245Wb.

[0030] Step 1 specifically involves: building a mathematical model of the hub motor and transforming the coordinates of the feedback signal.

[0031] Step 11: Establish the mathematical model of the permanent magnet synchronous motor.

[0032] Assuming the hub motor rotor has no damped windings, the magnetic circuit is unsaturated, core losses and eddy current effects are ignored, the three-phase windings are symmetrically distributed, and the induced electromotive force of the phase windings is sinusoidal, then the stator voltage equations in the synchronous rotating dq coordinate system are as shown in equations (1) and (2): (1) (2) In the formula, , These are the d-axis and q-axis stator voltages, respectively. , These are the d-axis and q-axis stator currents, respectively. Stator resistance; , These are the d-axis and q-axis inductances, respectively. For permanent magnet flux linkage; ω is the rotor's electric angular velocity.

[0033] The electromagnetic torque equation and the mechanical motion equation are shown in equations (3) and (4): (3) (4) In the formula, Electromagnetic torque; It is the extreme logarithm; The moment of inertia referred to the motor shaft; This is the load torque; It is the coefficient of viscous friction; Let be the rotor's mechanical angular velocity, and = · .

[0034] Step 12: To meet the requirements of digital control, the stator voltage equation is discretized using the first-order Euler discretization method, resulting in the current prediction models shown in equations (5) and (6), which serve as the prediction models for the lower-level dual-vector model predicting the current loop. (5) (6) In the formula, The sampling period; , For the first Current sample values ​​at each sampling time; , For the first The predicted current value at each sampling time.

[0035] Step 13: Feedback signal coordinate transformation.

[0036] During each control cycle, the three-phase stator current of the hub motor is collected by a current sensor. , , Using the three-phase current symmetry condition, the current components in the two-phase stationary coordinate system are obtained by the Clark transformation shown in equation (7). , : (7) The rotor electrical angular position is then measured using a rotary transformer or encoder. The d-axis and q-axis feedback currents are obtained through the Park transformation shown in equation (8). , : (8) Rotor electric angular velocity Obtained by differentiation of position signals or by observation.

[0037] Step 2 specifically involves: building the upper-layer self-disruption rejection speed loop.

[0038] Step 21: Establish the total disturbance model of the speed loop.

[0039] Based on the mechanical equations of motion, the load torque, viscous friction, and unmodeled disturbances are considered as the total disturbance of the system. As shown in equation (9): (9) The rotational speed ring can then be written in the standard form of a first-order integral series, as shown in equation (10): (10) In the formula, The torque coefficient, To control the gain. Due to the viscous friction coefficient Since the impact is closely related to the working environment and is of a relatively low magnitude, it can be included in the total disturbance for observation and compensation, without the need for precise modeling.

[0040] Step 22: Construct an improved extended state observer.

[0041] Total disturbance Expand the variables to new state variables and construct a second-order expanded state observer. The perturbation observations in the standard expanded observer... Only observation error The time integration results in low utilization of error signals and slow observation speed for sudden disturbances. Therefore, this embodiment adds a proportional feedback branch for observation errors to the disturbance observation channel, constructing the improved extended state observer shown in equations (11) to (13), with the structure as follows: Figure 3 As shown: (11) (12) (13) In the formula, This refers to the actual measured speed of the motor. These are observed rotational speed values; This represents the total observed disturbance of the system. This is the error in the rotational speed observation; , This is the observer error feedback gain coefficient; This is the proportional feedback gain coefficient for the rotational speed observation error. The branch introduces proportional feedback of observation error into the disturbance observation channel. Its equivalent effect is to contribute part of the speed feedback gain to the disturbance observation channel, thereby reducing the observation pressure on the integration channel and accelerating the disturbance observation speed without reducing the observation accuracy.

[0042] The characteristic equation of the observer coefficient matrix is ​​shown in equation (14): (14) To ensure the stability of the observer, , The values ​​must be positive real numbers; to simplify parameter design, the eigenvalues ​​are uniformly configured within the observation bandwidth. At this point, the parameter tuning relationship is shown in equation (15): (15) and Not greater than This is to avoid disturbing the observations and causing high-frequency fluctuations. During engineering tuning, the observation bandwidth should be used first. Sure , Increase it step by step The observation rate of the disturbance is maintained until the observation value does not show significant high-frequency fluctuations.

[0043] Step 23: Nonlinear error feedback law and overcurrent limiting.

[0044] To balance response speed and steady-state accuracy, the nonlinear fal function shown in equation (16) is used to construct the error feedback law. This error feedback law is the nonlinear state error feedback control law, abbreviated as NLSEF, which has the characteristics of large error and small gain, and small error and large gain. (16) In the formula, The difference between the setpoint speed and the observed speed; 0 < <1; The threshold value for the linear segment interval is used. The output of the error feedback law is compensated by the total disturbance observations to obtain the unlimited setpoint value of the quadrature-axis current. As shown in equation (17): (17) The overcurrent limiting circuit, as shown in equation (18), outputs the quadrature axis current setpoint. : (18) In the formula, Given a rotational speed; The feedback gain coefficient of the error feedback law; This represents the maximum permissible quadrature-axis current of the motor. Overcurrent limiting prevents damage to power devices due to overcurrent; the direct-axis current setpoint is [value to be inserted here]. =0, using Rotor field orientation control mode with =0.

[0045] Step 3 specifically involves: constructing the lower-level dual-vector model predictive current loop, the control flow of which is as follows: Figure 4 As shown.

[0046] Step 31: Inverter voltage vector description.

[0047] Two-level three-phase voltage source inverter topology as follows Figure 7 As shown, the three-channel inverter bridge has 8 switching states, corresponding to 6 effective voltage vectors and 2 zero voltage vectors, for a total of 7 basic voltage vectors; represented by switching quantities... , , These represent the switching states of the upper transistors of the three-phase bridge arms. When the switching quantity is 1, the corresponding switching element is turned on. The two switching elements on the same bridge arm cannot be turned on at the same time. The relationship between the output voltage vector and the switching state is shown in equation (19): (19) The relationship between the motor phase voltage and each switch state quantity is shown in equation (20): (20) In the formula, It is the DC bus voltage, supplied by the vehicle's power battery.

[0048] Step 32: First voltage vector selection.

[0049] The seven basic voltage vectors are transformed by Park to obtain the corresponding d-axis and q-axis voltage components. These components are then substituted into the current prediction model equations (5) and (6) to predict the d-axis and q-axis currents at the next sampling time. Finally, these components are substituted into the value function shown in equation (21). (twenty one) Iterate through and compare the value function values ​​corresponding to each basic voltage vector, and select the value that makes the value equal to ... The minimum voltage vector is the first optimal voltage vector. .

[0050] Step 33: Voltage vector action time allocation.

[0051] Will With 7 candidate voltage vectors Each voltage vector is combined separately, and the duration of action of the two voltage vectors in each combination within one sampling period is pre-allocated. The q-axis current deadbeat principle is adopted, letting... effect , effect( The q-axis current then reaches the given value at the next sampling time, as shown in equation (22): (twenty two) Solving The duration of action is shown in equation (23): (twenty three) And limit its amplitude to [0, according to formula (24)] ]: (twenty four) The slopes of the q-axis current changes under the action of each voltage vector are shown in equations (25) and (26): (25) (26) In the formula, The slope of the q-axis current change when a zero voltage vector is applied; , They are respectively and The corresponding q-axis component of the stator voltage.

[0052] Step 34: Second voltage vector selection.

[0053] Considering the duration of action of the two voltage vectors, the equivalent d-axis and q-axis voltages are weighted and synthesized according to their duration, as shown in equation (27): (27) Substituting the values ​​into the current prediction model yields the predicted current values ​​under seven vector combinations. Then, substituting these values ​​into the value function calculates the evaluation value. The candidate vector that minimizes the value function is selected as the second optimal voltage vector. . The predicted value of the electromagnetic torque at time t is shown in equation (28): (28) Since the second voltage vector can be a non-zero voltage vector during the second vector selection, the synthesized voltage vector can be extended to any direction and any amplitude, such as... Figure 5 As shown, the value function considers the action time of the voltage vector while pre-selecting it, avoiding the problem of the selected vector being no longer optimal due to selecting the vector first and then calculating the duty cycle in duty cycle MPCC, thus making the voltage vector selection more accurate. A total of 14 current predictions and value function evaluations are performed in each sampling period.

[0054] Step 35: Two-vector pulse synthesis.

[0055] The waveform generation module is in the first sampling period Output within time The corresponding three-phase switch states, in the remaining ( Switch output within a time period The corresponding three-phase switch states generate six switching pulse signals. ~ The pulse signals of the two switches on the same bridge arm are complementary and dead-time protection is added. The pulse synthesis timing is as follows: Figure 6 As shown, the pulse signal is amplified by the drive circuit and drives the power switching transistor of the three-phase voltage source inverter to output voltage to the stator winding of the hub motor.

[0056] Step 4 specifically involves setting evaluation indicators for torque fluctuation suppression effect.

[0057] The torque ripple coefficient shown in equation (29) is adopted. : (29) In the formula, For the steady-state maximum electromagnetic torque, It is the minimum electromagnetic torque in steady state.

[0058] The current harmonic distortion rate (THD) shown in equation (30) is the ratio of the amplitude of all harmonic content to the amplitude of the fundamental frequency: (30) In the formula, The fundamental amplitude, The amplitude of each harmonic is given.

[0059] The longitudinal impact jerk shown in equation (31) is the rate of change of the longitudinal acceleration of the entire vehicle: (31) During the Tip-in / out process (rapid acceleration / rapid release of the accelerator pedal), the longitudinal oscillation of the entire vehicle is caused by the torque fluctuation of the hub motor, which is described by the longitudinal impact.

[0060] and the rotational speed fluctuation shown in equation (32) Recovery time: Under conditions of sudden load torque application and removal, the maximum deviation of the measured speed from the given speed is taken as the speed fluctuation. (32) The recovery time is used as the time required for the speed to re-enter the error band of the given value ±2%, and the ability of the upper-level active disturbance rejection speed loop to suppress load disturbances is evaluated. The torque fluctuation suppression effect of the present invention is evaluated from four levels: torque, current, vehicle speed, and speed.

[0061] Example 2 This embodiment verifies the torque fluctuation suppression effect of the method described in Embodiment 1. It combines a disturbance observation comparison experiment between the improved extended state observer and the standard extended state observer, a disturbance rejection performance comparison experiment between three speed loop control methods, and a simulation comparison between the two-layer control scheme and the traditional scheme. The simulation is performed in the Matlab / Simulink environment. Figure 2 The simulation model and experimental platform were built based on the framework shown. The experimental conditions were as follows: a given rotational speed of 300 r / min was applied, and a load torque of 5 N·m was suddenly applied through the loading platform. After the rotational speed stabilized, the load was suddenly removed. The observation bandwidth of the observer was set to... =100, error feedback gain =0.03.

[0062] (1) Comparative experiment of perturbation observation between the improved extended state observer and the standard extended state observer.

[0063] Under the same observation bandwidth, the improved extended state observer with proportional feedback of rotational speed observation error, as shown in equations (11) to (13), and the one without... Standard extended state observer of the branch (i.e., let) =0), and the disturbance observation process of sudden application and sudden removal of 5 N·m load torque is compared. The disturbance observation curve is as follows: Figure 8 As shown in Table 1, the disturbance observation time and steady-state observation error under different observation bandwidths are as follows.

[0064] Table 1 Comparison of disturbance observation performance of the two extended state observers

[0065] Table 1 shows that, under the three observation bandwidths, the improved extended state observer has shorter rise and fall times than the standard extended state observer. The speed-up effect is most significant when the observation bandwidth is small, with rise and fall times reduced from 940ms and 760ms to 740ms and 540ms respectively, representing a speed-up of approximately 21%–29%. At the optimal observation bandwidth… At 100, the observed rise and fall times were shortened from 760ms and 300ms to 700ms and 260ms, respectively, and the steady-state observation error was comparable to that of the standard extended state observer (-1.0% vs. -1.4%), indicating that... The branch accelerates the observation speed under disturbances without reducing observation accuracy. When the observation bandwidth increases to 1000, the observation speed improvement of both observers tends to saturate, but the fluctuation of the observed values ​​intensifies; while when the observation bandwidth is small, the steady-state observed value of the improved extended state observer has a certain overshoot. Therefore, this embodiment optimizes the observation bandwidth. =100.

[0066] (2) Comparison of the anti-disturbance performance of three speed loop control methods.

[0067] PI controller and active disturbance rejection controller with standard extended state observer are used respectively. =0) and an active disturbance rejection controller with an improved extended state observer constitute the speed loop. The lower layer uses a dual-vector model to predict the current loop. The speed disturbance rejection performance is compared under sudden addition and removal of a 5 N·m load torque. The speed response curves are as follows: Figure 9 As shown in Table 2, the experimental data are as follows.

[0068] Table 2. Experimental data on disturbance rejection performance of three speed loop control methods

[0069] Table 2 shows that the system's disturbance rejection performance is significantly better than that of traditional PI control after adopting the active disturbance rejection speed loop. Furthermore, compared to the active disturbance rejection controller using the standard extended state observer, the speed fluctuation during sudden loading is further reduced from 12 r / min to 8 r / min, a reduction of approximately 33%, and the speed fluctuation during sudden unloading is reduced from 29 r / min to 18 r / min, a reduction of approximately 38%. The recovery times are also shortened from 380 ms and 220 ms to 360 ms and 180 ms, respectively. This is due to... The branch circuit accelerates the observation speed of sudden disturbances, enabling the disturbance compensation current to be injected more promptly, thereby further reducing speed fluctuations and consequently reducing electromagnetic torque fluctuations.

[0070] (3) Comparison and simulation of the two-layer control scheme and the traditional scheme.

[0071] Compared with the traditional scheme of PI speed loop + duty cycle MPCC current loop, simulation results show that under steady-state conditions, after adopting the dual-vector model to predict the current loop, the d-axis current fluctuation and stator current harmonic distortion rate are significantly reduced, and the electromagnetic torque fluctuation coefficient is significantly decreased. Under transient conditions, when the load torque is suddenly applied or removed, the speed fluctuation using the self-disturbance rejection speed loop is reduced by more than 60% compared with the PI speed loop, the recovery time is significantly shortened, and the speed step response has almost no overshoot. The electromagnetic torque transition process is smoother, and the longitudinal impact of the whole vehicle is significantly reduced. The synergistic effect of the upper self-disturbance rejection speed loop and the lower dual-vector model to predict the current loop effectively suppresses the output torque fluctuation of the hub motor, verifying the effectiveness and feasibility of the method of the present invention.

[0072] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for suppressing torque fluctuations in a hub motor system based on dual-layer control, characterized in that, A two-layer control architecture is adopted, consisting of an upper-layer active disturbance rejection speed loop and a lower-layer dual-vector model predictive current loop; The upper-level active disturbance rejection speed loop treats the load torque, viscous friction, and unmodeled dynamics as the total system disturbance and expands them into new state variables. An improved extended state observer with proportional feedback of speed observation error is constructed to observe the motor speed and the total system disturbance in real time. The difference between the speed setpoint and the speed observation value is calculated by the nonlinear fal function error feedback law, and the total disturbance observation value is converted into a compensation current for feedforward compensation. After current limiting, the quadrature axis current setpoint is output. The lower-level dual-vector model predicts the current loop by selecting two optimal voltage vectors in each sampling period based on the quadrature-axis current setpoint, the direct-axis current setpoint, and the quadrature-axis and direct-axis current feedback values, and allocating their action time. The pulse signal is then synthesized by the waveform generation module to drive the three-phase voltage source inverter to control the hub motor.

2. The torque fluctuation suppression method for a hub motor system based on dual-layer control according to claim 1, characterized in that, A mathematical model of a permanent magnet synchronous motor for hub motors in a synchronously rotating dq coordinate system is established, including stator voltage equation, electromagnetic torque equation, and mechanical motion equation. The stator voltage equation is discretized using the first-order Euler discretization method to obtain a current prediction model. Three-phase stator currents are collected, and the current components in the two-phase stationary coordinate system are obtained through Clark transformation. Then, the d-axis and q-axis feedback currents are obtained by combining the rotor electrical angular position with Park transformation.

3. The torque fluctuation suppression method for a hub motor system based on dual-layer control according to claim 1, characterized in that, The improved extended state observer expands the total system disturbance into new state variables, and the speed observation value is obtained by the speed observation error through the error feedback gain coefficient. Feedback correction: The total system disturbance observation value is calculated by adjusting the speed observation error via the error feedback gain coefficient. Feedback and proportional feedback gain coefficient Proportional feedback is used for joint correction.

4. The torque fluctuation suppression method for a hub motor system based on dual-layer control according to claim 3, characterized in that, The eigenvalues ​​of the characteristic equation of the coefficient matrix of the improved extended state observer are uniformly configured in the observation bandwidth. At this point, the error feedback gain coefficient , All are positive real numbers, and the proportional feedback gain coefficient Not greater than the error feedback gain coefficient .

5. The torque fluctuation suppression method for a hub motor system based on dual-layer control according to claim 1, characterized in that, The nonlinear FAL function error feedback law adopts a nonlinear FAL function with the characteristics of large error and small gain, and small error and large gain. The parameters of the nonlinear FAL function satisfy 0 < <1, The threshold for the linear segment interval; The output of the error feedback law is compensated by the total disturbance observation value to obtain the quadrature axis current unlimited setpoint value. After passing through the current limiting circuit, the quadrature axis current setpoint value is output, and the direct axis current setpoint value is zero.

6. The torque fluctuation suppression method for a hub motor system based on dual-layer control according to claim 1, characterized in that, The lower-level dual-vector model predicts the current loop by selecting voltage vectors twice in each sampling period. The candidate vectors for the second vector selection are not limited to the zero voltage vector. The action time of the two voltage vectors is allocated based on the q-axis current deadbeat principle. The value function considers the action time of the voltage vectors while pre-selecting them. Finally, the pulse is synthesized by the waveform generation module to drive the three-phase voltage source inverter.

7. The torque fluctuation suppression method for a hub motor system based on dual-layer control according to claim 6, characterized in that, The first voltage vector selection process involves: Taking the seven basic voltage vectors (six effective voltage vectors and one zero voltage vector) from the two-level three-phase voltage source inverter, performing a Park transform to obtain the corresponding d-axis and q-axis voltage components. These components are then substituted into the current prediction model to predict the d-axis and q-axis currents at the next sampling time. Finally, the values ​​of each basic voltage vector are compared using the value function, and the voltage vector that minimizes the value function is selected as the first optimal voltage vector. .

8. The torque fluctuation suppression method for a hub motor system based on dual-layer control according to claim 7, characterized in that, The specific time allocation for the voltage vector action is as follows: the first optimal voltage vector... Combined with 7 candidate voltage vectors, the effective time of two voltage vectors in each combination within one sampling period is allocated using the q-axis current deadbeat principle. Let the first optimal voltage vector be... effect Candidate voltage vector action - The q-axis current reaches the given value at the next sampling time, thus obtaining the first optimal voltage vector. The duration of action is limited to [0, ].

9. The torque fluctuation suppression method for a hub motor system based on dual-layer control according to claim 8, characterized in that, The second voltage vector selection is as follows: Considering the duration of action of two voltage vectors, the equivalent d-axis and q-axis voltages are weighted according to their duration and synthesized. This weighted sum is then used in the current prediction model to predict the current values ​​under seven vector combinations. The corresponding value function values ​​are calculated by substituting these values ​​into the value function, and the candidate vector that minimizes the value function is selected as the second optimal voltage vector. ; The waveform generation module first outputs the first optimal voltage vector within one sampling period. The corresponding three-phase switch state and its duration of action. Then switch to output the second optimal voltage vector. The corresponding three-phase switch status and continue - It generates six switching pulse signals to drive a three-phase voltage source inverter, with the pulse signals of the two switching transistors on the same bridge arm being complementary.

10. The torque fluctuation suppression method for a hub motor system based on dual-layer control according to claim 1, characterized in that, The torque fluctuation suppression effect is evaluated using torque fluctuation coefficient, current harmonic distortion rate, longitudinal impact, speed fluctuation amount and recovery time. Among them, current harmonic distortion rate is the ratio of the amplitude of all harmonic content to the amplitude of the fundamental wave, longitudinal impact is the rate of change of the longitudinal acceleration of the whole vehicle, speed fluctuation amount is the maximum deviation of the measured speed from the given speed under the condition of sudden load torque increase and decrease, and recovery time is the time required for the speed to re-enter the error band of the given value ±2%.