A robust multi-objective model predictive control method for AC permanent magnet synchronous motors

Abstract

The present invention discloses a robust multi-objective model predictive control method for an AC permanent magnet synchronous motor. This method is used in the control system of an AC permanent magnet synchronous motor and includes the following steps: collecting the operating parameters of the control system at time k; obtaining an electromagnetic torque reference value and a stator flux reference value; loading a multi-objective ranking model, and obtaining the electromagnetic torque ranking value and the stator flux ranking value based on the operating parameters at time k, after disturbance compensation using an improved full-order state observer and two-step delay compensation; the electromagnetic torque ranking value and the stator flux ranking value form a ranking value; outputting an optimal vector voltage; and achieving torque control through the vector voltage. This technical solution effectively improves system robustness, reduces computational complexity, reduces torque ripple, and improves control accuracy and the response speed of the MPTC.

Description

Technical Field

[0001] The present invention relates to the field of motor control, and in particular to a robust multi-objective model predictive control method for an AC permanent magnet synchronous motor. Background Art

[0002] Permanent magnet synchronous motors (PMSMs) offer advantages such as low energy consumption and stable operation, and are considered a green alternative to gasoline engines. Motor control is the core technology of PMSM electric drive systems, making the research of high-performance motor control algorithms crucial. With the rapid development of industrial technology, finite control set model predictive control (FCS-MPC), which easily implements multi-objective control and multiple constraints, has gradually entered the motor control field. There are two approaches to model predictive control (MPC) for AC PMSMs: field-of-control (FOC) and direct torque control (DTC). FOC decouples flux from torque through a series of coordinate transformations, then uses a PI controller to independently control the speed and current components to achieve the desired control objectives. However, its model is relatively simple, with few adjustable parameters, making it easy to control but difficult to tune. DTC is a high-performance AC variable frequency control technology that does not rely on complex coordinate transformations or pulse width modulation. Instead, it directly controls flux and torque through a hysteresis controller, selecting a voltage vector to control motor operation based on the error. This method is simple, practical, and easy to implement, but it is limited by the hysteresis width and sampling frequency, and suffers from problems such as large torque pulsation and excessive switching frequency, which restrict its application in high-performance scenarios. In recent years, due to the rapid development of digital technology and hardware equipment, model predictive control (MPC) has gained considerable attention in the field of motor drives due to its high dynamic performance, simple and easy-to-implement algorithms, and flexible value function design for multivariable tracking control. Combining the advantages of MPC and DTC, model predictive torque control (MPTC) directly servo-tracks torque and flux, omitting the hysteresis controller and replacing the switching table with rolling optimization. It has good dynamic response performance, smaller torque pulsation than DTC, and more precise control. Therefore, it is widely used in torque and flux prediction scenarios.

[0003] However, the MPTC model is highly dependent on parameters. In the actual operation of the motor, due to parameter measurement errors, changes in external environmental factors and external environmental interference, excessive torque pulsation, and model parameter mismatch, the system has poor robustness and is prone to parameter mismatch, resulting in poor prediction accuracy and control effect. Figure 1As shown in the figure, resistor mismatch (e.g., resistance R = 2R) affects the MPTC model's predictions. To address this issue, other intelligent algorithms, such as Kalman filtering, sliding film observers, and least squares methods, can be used for parameter identification. However, this increases the system's computational burden and results in lower control accuracy compared to ESO. For example, other intelligent algorithms, such as multi-objective genetic algorithms and neural networks, can replace the sorting concept, but this increases the system's computational burden and reduces the MPTC's response speed.

[0004] Therefore, a technical solution is needed to achieve accuracy that is independent of parameters, reduce sensitivity to parameter changes, and achieve dynamic response of predicted torque. Summary of the Invention

[0005] To achieve the above objectives, the present application provides a robust multi-objective model predictive control method for an AC permanent magnet synchronous motor. The method is used in a control system of an AC permanent magnet synchronous motor and includes the following steps:

[0006] Collect the operating parameters of the control system at time k, including: the electrical angular velocity of the motor rotor , rotor position θ, three-phase motor current 、 and , DC bus voltage and the d and q axis components of the current 、 ;

[0007] Obtain electromagnetic torque reference value through PI speed outer loop , obtain the stator flux reference value through the maximum torque current ratio (MTPA) ;

[0008] Load the multi-objective sorting model and obtain the electromagnetic torque sorting value based on the operating parameters at time k, the electromagnetic torque and stator flux at time k And the sort value of the magnetic link ;Electromagnetic torque ranking value And the sort value of the magnetic link Composition sort value ; Obtain and output optimal vector voltage ; Through the vector voltage Achieve torque control;

[0009] The multi-objective sorting model includes: observing and compensating the current error in real time through the improved ESO observer; equating the disturbance of parameter mismatch to an independent input quantity and defining the disturbance; performing two-step delay compensation to obtain the stator flux at time k+2. and ; Reconstruct the value function and output the electromagnetic torque ranking value And the sort value of the magnetic link .

[0010] Among them, the observation and real-time compensation of current errors caused by external parameter disturbances include:

[0011] Adding external disturbances to the state equation of PMSM in the synchronous rotating coordinate system is expressed as:

[0012]

[0013] in, and is the system voltage input on the d and q axes, is the external disturbance value other than resistance, inductance and magnetic flux, 、 、 All are nominal values ​​on the motor nameplate. 、 、 are the perturbation values ​​of each parameter.

[0014] The disturbance of parameter mismatch is equivalent to a new input definition disturbance, which can be expressed as:

[0015]

[0016] in, 、 is the total system disturbance on the d and q axes, 、 、 is the nominal value on the motor nameplate, 、 、 are the perturbation values ​​of each parameter, are external disturbance values ​​other than resistance, inductance, and magnetic flux.

[0017] After defining the perturbation, the spatial state equation can be reformulated as:

[0018]

[0019] at this time, and is the system voltage input on the d and q axes, and the interference term , The observed value of can be obtained through the expression of the second-order ESO.

[0020] Current and Perform two-step delay compensation to obtain the stator flux at time k+2 and electromagnetic torque , get the torque error at time k+2 and flux linkage error , the calculation method is:

[0021] ,

[0022] ,

[0023] ,

[0024] ,

[0025] Furthermore, the reconstruction value function is used to obtain the electromagnetic torque ranking values ​​by linear combination. And the sort value of the magnetic link , expressed as:

[0026] ,

[0027] Furthermore, the 8 voltage vectors can be reduced to 4 voltage vectors through the graph theory algorithm.

[0028] Expressed as: ,

[0029] Among them, the optimal vector voltage is expressed as: .

[0030] The robust multi-objective model predictive control method provided by the present invention takes into account the changes in motor parameters during actual motor operation, which affect the system control performance. The addition of an ESO observer can effectively improve the system robustness, reduce the amount of calculation, and reduce torque ripple. At the same time, the current of the ESO observer is quadratically compensated, the cost function formula is rewritten, and the idea of ​​sorting is used instead of determining the weight coefficient. This not only improves the control accuracy, but also increases the response speed of the MPTC. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a diagram showing the changes in speed, torque and flux linkage under resistance mismatch conditions according to the prior art of the present invention;

[0032] Figure 2 is a step diagram of a method for predicting torque control through a multi-objective ranking model provided in accordance with an embodiment of the present invention;

[0033] Figure 3 This is a comparison diagram of the effects of torque control achieved by the method provided by an embodiment of the present invention and that achieved by traditional MPTC;

[0034] Figure 4 This is a comparison diagram of the effects of speed control achieved by the method provided by an embodiment of the present invention and that achieved by traditional MPTC. DETAILED DESCRIPTION

[0035] In order to improve the current error and oscillation problems of permanent magnet synchronous motors caused by parameter mismatch and external disturbances, the present invention designs an improved extended observer (ESO) on the basis of single vector model predictive torque control to observe and compensate for the motor parameter disturbances in real time, thereby achieving the effect of suppressing steady-state prediction errors and improving the robustness of the system. In order to solve the delay caused by the computer instruction cycle, two-step delay compensation is performed in the calculation process to obtain the final stator flux and electromagnetic torque. In the present invention, the cost function is also optimized for the calculation results after the two-step delay compensation, and finally the electromagnetic torque and flux are obtained, and the optimal voltage vector is output.

[0036] The specific implementation of the present invention is described in detail below with reference to the accompanying drawings.

[0037] Figure 2 A step diagram of a robust multi-objective model predictive control method for a control system of an AC permanent magnet synchronous motor implemented by the present invention is provided, as shown in the figure, comprising the following steps:

[0038] Step S100: The control system collects the operating parameters corresponding to the k-th moment, including: the electrical angular velocity of the motor rotor , rotor position θ, three-phase motor current 、 and , DC bus voltage , and solve for the d and q axis components of the current 、 .

[0039] Step S110: Obtaining electromagnetic torque reference value , stator flux reference value ;

[0040] Calculate the speed error through the PI speed outer loop , and obtain the electromagnetic torque reference value ,

[0041] The stator flux reference value is obtained by the torque current ratio (MTPA) ;

[0042] Step S120: Load the multi-objective sorting model and obtain the electromagnetic torque sorting value according to the operating parameters at time k, the electromagnetic torque and the stator flux at time k. And the sort value of the magnetic link ; Sort by electromagnetic torque value And the sort value of the magnetic link Composition sort value ;

[0043] In general, the traditional calculation model can be based on the d and q axis components 、 , use the torque formula to get the electromagnetic torque at time k and magnetic links , the specific calculation formula is as follows:

[0044]

[0045] in, is the permanent magnet flux linkage value, is the d-axis inductance value and the q-axis inductance value, in the surface-mounted permanent magnet synchronous motor .

[0046] However, this calculation model does not take into account the situation of parameter mismatch and interference between parameters.

[0047] Based on this, the present invention adopts a multi-objective ranking model for improvement:

[0048] Step S121: Design an improved ESO observer to observe and compensate for current errors caused by external parameter disturbances in real time.

[0049] The state space equation of PMSM in the synchronous rotating coordinate system is as follows:

[0050]

[0051] In the present invention, the voltage equation is rewritten as follows, taking into account parameter mismatch:

[0052]

[0053] in, is the voltage of the d-axis and q-axis input to the system, and These are various parameter disturbances and external interferences on the d-axis and q-axis.

[0054] Step S122: Based on the ESO observation concept, the disturbance caused by parameter mismatch is equivalent to a new input quantity, that is, the ESO definition disturbance is as follows:

[0055]

[0056] in, , is the total disturbance of the system on the d and q axes, , , is the nominal value on the motor nameplate, , , are the perturbation values ​​of each parameter, are external disturbance values ​​other than resistance, inductance, and magnetic flux.

[0057] At this point, the spatial state equation of the surface-mounted permanent magnet synchronous motor can be re-expressed as:

[0058] (7)

[0059] at the same time, and is the system voltage input on the d and q axes, and the interference term is obtained from the expression of the second-order ESO , Observations of:

[0060] Taking the d-axis as an example, the current The expression of the second-order ESO is as follows:

[0061]

[0062] The nonlinear error feedback control law is:

[0063]

[0064] Where, It is a piecewise function, which will produce chattering phenomenon at the switching point. The sliding membrane control system (SMC) can achieve the function of stable tracking of the target by controlling the sliding of the system state on the sliding surface, which can eliminate the chattering phenomenon of the traditional ESO, that is, the improved The function is:

[0065]

[0066] Next, the improved ESO is discretized, taking the d-axis as an example:

[0067] (11)

[0068] in, is the gain of the discrete ESO.

[0069] Step S123: Perform two-step delay compensation:

[0070] In the present invention, the current and Perform two-step delay compensation and use formulas 1, 2, and 3 to obtain the stator flux at time k+2. and electromagnetic torque , get the torque error at time k+2 and flux linkage error , the calculation method is:

[0071] (12)

[0072] (13)

[0073] (14)

[0074] (15)

[0075] Step S124: Reconstruct the value function and use linear combination to obtain the electromagnetic torque ranking values And the sort value of the magnetic link , expressed as:

[0076] (16)

[0077] Furthermore, the graph theory algorithm can reduce the 8 voltage vectors to 4 voltage vectors, which can be expressed as:

[0078] (17)

[0079] Sort value Minimum voltage Expressed as:

[0080] (18)

[0081] Step S130: Obtain and output the optimal vector voltage; in the present invention, as in formula 18, the ranking value Minimum voltage , this voltage value is the optimal voltage vector, which is used to realize the permanent magnet synchronous motor model predictive control.

[0082] The present invention provides a simulation comparison of torque control prediction by multi-objective sorting model under certain working conditions, and the torque effect is as follows: Figure 3 As shown, the speed effect is as follows Figure 4 According to the simulation results, the improved ESO model predictive torque control has significantly reduced torque fluctuations and faster speed response compared to the traditional model predictive torque control.

[0083] The robust multi-objective model predictive control method provided by this invention, based on single-vector model predictive torque control, observes and compensates for motor parameter disturbances in real time by designing an improved extended observer (ESO), thereby suppressing steady-state prediction errors and improving system robustness. Furthermore, considering the impact of computer instruction cycle delays, two-step delay compensation is performed to obtain the final stator flux and electromagnetic torque, which are then substituted into the optimized cost function to support the output of the optimal voltage vector. This method differs from traditional MPTC in that it separately considers disturbances affecting motor parameters in actual operating scenarios and incorporates them into the ESO observer, effectively improving system robustness while reducing computational effort and torque ripple. Furthermore, quadratic compensation of the ESO observer's current improves not only its control accuracy but also the MPTC's response speed.

[0084] The above disclosures are only a few specific embodiments of the present invention. However, the present invention is not limited thereto. Any changes that can be conceived by those skilled in the art should fall within the scope of protection of the present invention.

Claims

1. A robust multi-objective model predictive control method for an AC permanent magnet synchronous motor, characterized in that: A control system for an AC permanent magnet synchronous motor includes the following steps: Collect the operating parameters of the control system at time k, including: motor rotor electrical angular velocity , rotor position θ, three-phase motor current 、 and , DC bus voltage and the d and q axis components of the current 、 ; Obtain electromagnetic torque reference value through PI speed outer loop , obtain the stator flux reference value through the maximum torque current ratio MTPA ; Load the multi-objective sorting model and obtain the electromagnetic torque sorting value based on the operating parameters at time k, the electromagnetic torque and stator flux at time k And the sort value of the magnetic link ;Electromagnetic torque ranking value And the sort value of the magnetic link Composition sort value ; Obtain and output optimal vector voltage ; Through the vector voltage Achieve torque control; The multi-objective sorting model includes: observing and compensating the current error in real time through the improved ESO observer; equating the disturbance of parameter mismatch to an independent input quantity and defining the disturbance; performing two-step delay compensation to obtain the stator flux at time k+2. and ; Reconstruct the value function and output the electromagnetic torque ranking value And the sort value of the magnetic link ; Among them, the disturbance of parameter mismatch is equivalent to an independent input quantity, which is expressed as: in, 、 is the total system disturbance on the d and q axes, 、 、 is the nominal value on the motor nameplate, 、 are the perturbation values ​​of each parameter, Other external disturbance values ​​besides resistance, inductance and magnetic flux; After defining the perturbation, the spatial state equation is reformulated as: at this time, and is the system voltage input on the d and q axes, and the interference term 、 The observed value of can be obtained through the expression of the second-order ESO; The two-step delay compensation includes: The improved ESO is discretized, and the d-axis and q-axis are processed in the same way, where the d-axis is: ， in, is the gain of the discrete ESO, and: ; Current and Perform two-step delay compensation to obtain the stator flux at time k+2 and electromagnetic torque , get the torque error at time k+2 and flux linkage error , the calculation method is: ， ， ， ； The reconstructed value function is used to obtain the electromagnetic torque ranking values ​​respectively by using a linear combination method. And the sort value of the magnetic link , expressed as: ； The 8 voltage vectors are reduced to 4 voltage vectors, which can be expressed as: ; Among them, the optimal vector voltage is expressed as: .

2. The model predictive control method according to claim 1, characterized in that: in, Observing and real-time compensating for current errors include: Adding external disturbance to the state equation of PMSM in the synchronous rotating coordinate system is expressed as: in, 、 、 All are nominal values ​​on the motor nameplate. 、 、 are the perturbation values ​​of each parameter.

Images