CN-122001255-A - Permanent magnet synchronous motor model predictive control method and system

CN122001255ACN 122001255 ACN122001255 ACN 122001255ACN-122001255-A

Abstract

The invention relates to a permanent magnet synchronous motor model predictive control method and a permanent magnet synchronous motor model predictive control system, belongs to the technical field of permanent magnet synchronous motor driving, and solves the problem that in the prior art, when parameter mismatch occurs, the permanent magnet synchronous motor model predictive control cannot give consideration to both parameter robustness and dynamic performance. The method comprises the steps of obtaining stator current, stator voltage, sampling values of electric angular speed and direct current bus voltage at the current sampling moment and the previous two moments, updating time sequence sampling data of an increment current prediction error and the increment voltage, obtaining a first sample vector and a second sample vector at the current sampling moment, obtaining identification values of parameter mismatch coefficients at the current sampling moment by adopting a small batch random gradient descent method, obtaining stator current prediction values of different voltage vectors at the next moment, obtaining an optimal voltage vector, and performing permanent magnet synchronous motor model prediction control at the next moment based on the optimal voltage vector.

Inventors

ZHAO CHUQIAO
Mao yongle
CHEN XIN
QIN WENYUAN
QIAO HONGKAI
LIU ZHENGYU
YUE BINKAI

Assignees

北京机械设备研究所

Dates

Publication Date: 20260508
Application Date: 20241107

Claims (10)

1. The permanent magnet synchronous motor model prediction control method is characterized by comprising the following steps of: Acquiring sampling values of stator current, stator voltage and electric angular velocity at the current sampling moment and the first two moments and DC bus voltage at the current sampling moment, updating time sequence sampling data of an increment current prediction error and the increment voltage, and further acquiring a first sample vector and a second sample vector at the current sampling moment, wherein the time sequence sampling data of the increment current prediction error and the increment voltage are increment current prediction error and increment voltage calculated at each moment before the current moment; Based on the first sample vector and the second sample vector at the current sampling moment, a small batch random gradient descent method is adopted to obtain the identification value of the parameter mismatch coefficient at the current sampling moment, and then the stator current prediction value of different voltage vectors at the next moment is obtained; And obtaining an optimal voltage vector based on stator current prediction values of different voltage vectors at the next moment, and performing permanent magnet synchronous motor model prediction control at the next moment based on the optimal voltage vector.
2. The permanent magnet synchronous motor model predictive control method according to claim 1, wherein the first sample vector and the second sample vector at the current sampling time are obtained by: Based on the obtained sampling values and the electric angular speeds of the stator currents at the current sampling time and the previous two times, obtaining an H-axis current prediction error at the current time and the previous time, and further obtaining an H-axis incremental current prediction error at the current time; Obtaining H-axis increment voltage at the previous moment based on the obtained sampling values of the H-axis stator voltage at the previous two moments of the current sampling moment; The H-axis increment current prediction error at the current moment and the increment voltage at the previous moment are added into time sequence sampling data of the increment current prediction error and the increment voltage for updating; Sliding a sliding window constructed based on the set sliding window width to the latest sampling data position in the updated time sequence sampling data of the increment current prediction error and the increment voltage; Respectively taking the prediction error and the increment voltage of each H-axis increment current in the current sliding window as a first sample vector and a second sample vector at the current sampling moment; wherein the H axis is d axis or q axis.
3. The permanent magnet synchronous motor model predictive control method according to claim 1, wherein the identification value of the parameter mismatch coefficient at the current sampling time is obtained by: s211, selecting sampling data at corresponding moments in a first sample vector and a second sample vector at the current sampling moment based on the set small batch number to respectively form a first small batch sample vector and a second small batch sample vector of the current iteration; S212, obtaining a loss function value of a parameter mismatch coefficient in the current iteration based on the first small batch sample vector and the second small batch sample vector of the current iteration; S213, if the loss function value of the parameter mismatch coefficient in the current iteration meets the set error requirement, the parameter mismatch coefficient in the current iteration is used as the identification value of the parameter mismatch coefficient at the current moment, otherwise, the parameter mismatch coefficient in the current iteration is updated and used as the parameter mismatch coefficient of the next iteration, and the steps S211-S213 are repeated.
4. A permanent magnet synchronous motor model predictive control method according to claim 3, wherein the loss function of the parameter mismatch coefficient in the first iteration is expressed as: Where J (M est,l ) represents the loss function of the parameter mismatch coefficient M est,l in the first iteration, y b,l 、x b,l represents the first and second small sample vectors in the first iteration, respectively, Representing a binary norm.
5. The method for predictive control of a permanent magnet synchronous motor model according to claim 4, wherein the parameter mismatch coefficient of the next iteration is updated in the current iteration by: Where M est,l+1 represents the parameter mismatch coefficient in the 1+1st iteration, η represents the learning rate, Representing the gradient of the loss function J (M est,l ) of the parameter mismatch coefficient M est,l in the first iteration, x i representing the i-th sample data in the second small sample vector x b,l in the first iteration, and b representing the total number of sample data in the second small sample vector x b,l .
6. A permanent magnet synchronous motor model predictive control method according to claim 3, wherein the error requirement includes a loss function value being equal to or less than a set error threshold.
7. The permanent magnet synchronous motor model predictive control method according to claim 1, wherein the stator current predictive value of the different voltage vectors at the next time is obtained by: obtaining the current prediction error of d-axis and q-axis currents at the current moment based on the obtained current sampling moment, the sampling value of the current of the d-axis and the sampling value of the current of the stator of the q-axis at the previous moment and the electric angular velocity; Obtaining current prediction error compensation items of d-axis and q-axis at the current moment based on the current prediction errors of d-axis and q-axis at the current moment, the identification values of parameter mismatch coefficients and the obtained stator voltages of the d-axis and q-axis at the previous moment; Obtaining d and q axis voltage calculation values under different voltage vectors based on the obtained bus voltage at the current moment, wherein different voltage vectors are obtained based on different switching states of the inverter And obtaining the d-axis and q-axis stator current predicted values at different voltage vectors at the next moment based on the current predicted error compensation items of the d-axis and q-axis at the current moment, the stator current sampling values and the d-axis and q-axis voltage calculated values of the different voltage vectors.
8. The model predictive control method of a permanent magnet synchronous motor according to claim 7, wherein the d-axis and q-axis stator current predictive values at different voltage vectors at the next time are obtained by: Wherein i d,h (k+1)、i q,h (k+1) represents the predicted values of the d-axis and q-axis stator currents respectively at the h-th voltage vector at the k+1 time, Respectively representing d-axis and q-axis stator current sampling values at K time, u' d,h (k)、u′ q,h (K) respectively representing d-axis and q-axis stator voltage calculation values at h-th voltage vector at K time, L s representing stator inductance, R s representing stator resistance, T s representing sampling period, ω e (K) representing electrical angular velocity at K time, K d (k)、K q (K) respectively representing d-axis and q-axis current prediction error compensation terms at K time, ψ f representing permanent magnet flux linkage.
9. The model predictive control method for a permanent magnet synchronous motor according to claim 8, wherein the d-axis and q-axis current prediction error compensation terms K d (k)、K q (K) at the K-time are expressed as: Where M est (k) represents the parameter mismatch coefficient at time k, Respectively represent the d-axis and q-axis stator voltage sampling values at the time of k-1, The d-axis and q-axis current prediction errors at the k-time are shown, respectively.
10. A permanent magnet synchronous motor model predictive control system, comprising: The data acquisition and processing module is used for acquiring the stator current, the stator voltage and the sampling value of the electric angular velocity at the current sampling moment and the previous two moments and the DC bus voltage at the current sampling moment, updating the time sequence sampling data of the increment current prediction error and the increment voltage, and further obtaining a first sample vector and a second sample vector at the current sampling moment, wherein the time sequence sampling data of the increment current prediction error and the increment voltage are the increment current prediction error and the increment voltage calculated at each moment before the current moment; The stator current prediction value calculation module is used for obtaining the identification value of the parameter mismatch coefficient at the current sampling moment by adopting a small batch random gradient descent method based on the first sample vector and the second sample vector at the current sampling moment so as to obtain the stator current prediction value of different voltage vectors at the next moment, wherein different voltage vectors are obtained based on different switching states of the inverter; the model prediction control module is used for carrying out the model prediction control of the permanent magnet synchronous motor at the next moment based on the optimal voltage vector based on the stator current prediction values of different voltage vectors at the next moment.

Description

Permanent magnet synchronous motor model predictive control method and system Technical Field The invention relates to the technical field of permanent magnet synchronous motor driving, in particular to a permanent magnet synchronous motor model predictive control method and a permanent magnet synchronous motor model predictive control system. Background The surface-mounted permanent magnet synchronous motor is widely applied to the field of electric drive due to high efficiency, high power density and good control precision. With the technical progress of hardware processors, more complex control strategies are applied, and new control strategies with better control performance emerge in the field of motor driving. The model predictive control strategy has the advantages of simple structure, rapid dynamic response, compatibility with multi-target constraint processing and the like, and has good development prospect in the field of permanent magnet synchronous motor driving. However, in this strategy, the predictive model is built entirely on the basis of the mathematical model of the permanent magnet synchronous motor, so that the internal parameters of the controller must be highly matched to the actual parameters of the motor. In an actual working condition, internal parameters of the motor may change due to temperature rise, saturation effect and the like, so that mismatch occurs between the internal parameters of the controller and the actual parameters, and an incorrect optimal voltage vector is selected, so that serious current prediction errors are caused, and control performance is reduced. Therefore, improving the robustness of the parameters is particularly critical to exerting the advantages of the model predictive control strategy and guaranteeing the control performance of the model predictive control strategy. At present, three methods for improving the robustness are mainly three, namely a disturbance observer method, a model-free method and an on-line parameter identification method. However, the three methods have the defects that the observer has complex structural design, the disturbance estimation accuracy is influenced by multi-parameter setting and adjustment, the application of the observer in the engineering field is limited, the requirements on the hardware sampling frequency and the hardware sampling precision are increased by the model-free method, the hardware implementation is not easy, the application of the online parameter identification method is wider, the underrank problem exists, the nonlinear system identification problem is not easy to solve, the identification method is easy to be influenced by measurement noise and current pulsation, and the control performance is poor especially when the dynamic working condition is met. Therefore, a method for compensating prediction errors of the permanent magnet synchronous motor model prediction control is needed, so that when parameter mismatch occurs, the motor has good control performance and parameter robustness under steady-state working conditions and dynamic working conditions. Disclosure of Invention In view of the above analysis, the embodiment of the invention aims to provide a permanent magnet synchronous motor model predictive control method and a system, which are used for solving the problem that the conventional permanent magnet synchronous motor model predictive control cannot achieve both parameter robustness and dynamic performance when parameters are not matched. In one aspect, the embodiment of the invention provides a permanent magnet synchronous motor model prediction control method, which comprises the following steps: Acquiring sampling values of stator current, stator voltage and electric angular velocity at the current sampling moment and the first two moments and DC bus voltage at the current sampling moment, updating time sequence sampling data of an increment current prediction error and the increment voltage, and further acquiring a first sample vector and a second sample vector at the current sampling moment, wherein the time sequence sampling data of the increment current prediction error and the increment voltage are increment current prediction error and increment voltage calculated at each moment before the current moment; Based on the first sample vector and the second sample vector at the current sampling moment, a small batch random gradient descent method is adopted to obtain the identification value of the parameter mismatch coefficient at the current sampling moment, and then the stator current prediction value of different voltage vectors at the next moment is obtained; And obtaining an optimal voltage vector based on stator current prediction values of different voltage vectors at the next moment, and performing permanent magnet synchronous motor model prediction control at the next moment based on the optimal voltage vector. Preferably, the first sample vector and the second sample vec