Double-vector model predictive control method with current harmonic suppression

Abstract

The invention discloses a double-vector model prediction control method with current harmonic suppression. The method comprises the following steps: constructing a discrete current prediction model for double-vector model prediction; designing a sliding mode surface and selecting a reaching law, and designing a disturbance observer according to the sliding mode surface, the reaching law and the mathematical model; the output current of the disturbance observer is extracted under static coordinates; extracting fifth and seventh harmonics by using a frequency self-adaptive filtering controller, converting harmonic current into a rotating coordinate system, and performing feedback compensation on the harmonic current and originally output current to obtain current without the fifth and seventh harmonics; and injecting the current after harmonic wave filtering into the prediction model for modulation, and regulating and controlling the operation of the motor. According to the invention, the adverse effect of harmonic waves generated by the nonlinearity and dead zone effect of the inverter on the prediction precision of the motor control system is effectively reduced; a high-performance filtering effect is realized, and the system is endowed with high flexibility, so that the system can easily cope with various complex application scenes.

Description

Technical Field

[0001] The present invention relates to the technical field of permanent magnet synchronous motor control, and in particular to a dual-vector model predictive control method with current harmonic suppression. Background Art

[0002] Permanent magnet synchronous motor is a strongly coupled, complex nonlinear system. With its advantages of high power density, high reliability and high efficiency, it is widely used in new energy vehicles, wind power generation, aerospace, transportation and other fields. In the face of different control requirements, researchers have proposed a series of control algorithms in the hope that the motor control system can achieve the expected control performance. The most popular algorithms are: traditional PI vector control, direct torque control, sliding mode control and model predictive control, etc.

[0003] Although traditional PI vector control has advantages in adjusting the dynamic performance of the motor, it performs poorly in the face of disturbances such as parameter mismatch and speed change. Direct torque control has faster response speed and strong robustness, but the torque pulsation is large and the system is difficult to accurately control at low speed, and the stability performance is poor. In contrast, model predictive control combines the above two methods, with a simple structure, easy-to-understand principles, fast response, good robustness and easy implementation. At present, it has become a hot topic of research. Model predictive control can be divided into model predictive current control and model predictive torque control. In recent years, model predictive current control has been widely used in the study of high-performance control methods of permanent magnet synchronous motors. Specific control methods include: single vector, dual vector and triple vector, etc.

[0004] The traditional single-vector model predictive current control substitutes the eight basic voltage vectors generated by the two-level three-phase inverter into the current cost function, and selects an optimal voltage vector to act on the motor in one control cycle. However, selecting only one voltage vector in one cycle is prone to over-regulation or under-regulation, causing large current harmonics. For this reason, dual-vector control is introduced, and two voltage vectors are selected in one cycle, brought into the cost function, and the optimal voltage vector combination is selected. Both single-vector and dual-vector rely on accurate mathematical models of permanent magnet synchronous motors. If the model parameters do not match the actual situation, the predicted current will be inaccurate, affecting the system performance. In addition, nonlinear factors in the operation of the motor will also make the model parameters inconsistent with the actual situation. For this reason, a disturbance observer is designed for current compensation.

[0005] However, in the actual operation of the motor, due to the nonlinear characteristics of the inverter and the dead zone effect of the inverter, the harmonics of the current will be large, affecting the prediction accuracy of the control system. Therefore, considering how to reduce the interference caused by the high-order harmonics of the current in the permanent magnet synchronous motor to the motor operation control system has become a problem that needs to be solved. Summary of the invention

[0006] Purpose of the invention: The purpose of the present invention is to provide a dual-vector model predictive control method with current harmonic suppression.

[0007] Technical solution: The dual-vector model predictive control method with current harmonic suppression described in the present invention comprises the following steps:

[0008] (1) Based on the d-axis and q-axis voltage equations of the surface-mounted permanent magnet synchronous motor in the synchronous rotating coordinate system, a discretized current prediction model predicted by the dual-vector model is constructed;

[0009] (2) When parameter mismatch occurs, design the sliding surface and select the reaching law, and design the disturbance observer based on the sliding surface, reaching law and mathematical model;

[0010] (3) extracting the disturbance observer output current in a stationary coordinate system; extracting the fifth and seventh harmonics using a frequency adaptive filter controller, and transforming the harmonic current into a rotating coordinate system, performing feedback compensation with the original output current, and obtaining a current with the fifth and seventh harmonics removed;

[0011] (4) The current after harmonic filtering is injected into the prediction model for modulation, and the optimal voltage vector combination is selected to regulate the motor operation.

[0012] Furthermore, the voltage equations of the surface-mounted permanent magnet synchronous motor on the d and q axes in step (1) are as follows:

[0013]

[0014] Among them, u d with u q are the corresponding d-axis and q-axis voltages respectively; i d and i q is the d-axis and q-axis current; ω e is the electrical angular velocity; is the permanent magnet flux; R is the stator resistance, and the inductances of the d and q axes are approximately equal.

[0015] Furthermore, the process of constructing a discretized current prediction model predicted by the dual-vector model in step (1) includes:

[0016] The voltage equation can be rewritten as the current state equation as follows:

[0017]

[0018] Use the forward Euler discrete equation to predict the current i at the next moment d (k+1), iq(k+1), the current prediction equation is as follows:

[0019]

[0020] Among them, i d (k), i q (k) represents the actual current value at the current moment, i d (k+1), i q (k+1) is the predicted current value at the next moment; u d (k),u q (k) represents the basic voltage vector of the inverter applied at the next moment. In the dual vector model predictive control, u d (k),u q (k) can be expressed as:

[0021]

[0022] The first vector action time t1, according to the current zero beat principle i d =0, the calculation formula is as follows:

[0023]

[0024] Wherein, S1 and S2 are the current change rates when the first vector and the second vector act on each other.

[0025] Furthermore, the discretized current prediction model is used to predict the current generated by the synthetic voltage vector of two different vectors, using the cost function The optimal voltage vector combination with the smallest predicted current error is selected to perform current prediction and regulation.

[0026] Furthermore, the d-axis and q-axis voltage equations of the PMSM stator side with the disturbance error added in step (2) are expressed as follows:

[0027]

[0028]

[0029] Among them, f d 、f q represents the disturbance error caused by the d-axis and q-axis parameters, F d 、F q is the parameter f d 、f q The rate of change.

[0030] Furthermore, the disturbance observer is designed in step (2) as follows:

[0031]

[0032] in, denote the observed values ​​of the d-axis and q-axis currents, respectively, Respectively represent the observed values ​​of d-axis and q-axis disturbances; coefficient G d , G q represents the gain of sliding mode control; D dsmo , D qsmo Represents the sliding mode control function, and defines the current error and disturbance error as follows:

[0033]

[0034] Among them, e d 、e q is the error between the observed current value of d and q axis and the actual current value; e fd 、e fq is the error between the observed value of the d and q axis disturbance and the actual value of the disturbance. The observed value voltage equation and the actual value voltage equation are subtracted to obtain the error equation:

[0035]

[0036] Select e d 、e q As the sliding surface, that is: S d =e d , S q =e q , the sliding mode control function is designed using the constant velocity reaching law, and the expression is as follows:

[0037]

[0038] The reaching law parameter k is a positive value, and the error equation is rewritten as:

[0039]

[0040] Due to e fd , e fq is the disturbance quantity, the sampling cycle time is very short, and the error change rate of the parameter disturbance is considered to be 0. Substituting it into the above formula, the sliding mode control function is:

[0041]

[0042] According to the discretized equation of current, a reasonable G d , G q Value, the disturbance observer mathematical model is as follows:

[0043]

[0044] in, Respectively represent the observed values ​​of d-axis and q-axis disturbances; coefficient G d , G q represents the gain of sliding mode control; i d (k), iq (k) represents the actual current value at the current moment.

[0045] Furthermore, the current equations of the 5th and 7th harmonics in step (3) are as follows:

[0046]

[0047] Furthermore, the step (3) comprises:

[0048] The d-axis and q-axis currents output by the disturbance observer are transformed into the stationary coordinate system according to the inverse Park transformation to obtain i α (k+1), i β (k+1), the transformation formula is as follows:

[0049]

[0050] Harmonic extraction is performed using a frequency adaptive filter. According to the internal structure of the harmonic extraction unit, the transfer function of the frequency adaptive filter is as follows:

[0051]

[0052] Among them, ω0 represents the resonant frequency of the filter. According to the above formula, the input and output current transfer function based on the frequency adaptive filter is constructed:

[0053]

[0054] The parameter k determines the bandwidth of the filter; 5ω is injected into the frequency adaptive filter e 7ω e The resonant frequency is used to extract the 5th and 7th harmonic currents. The expressions are as follows:

[0055]

[0056] Extract the current containing the 5th and 7th harmonics in the stationary coordinate system: i α5 (k+1), i α7 (k+1), i β5 (k+1), i β7 (k+1);

[0057] Perform park transformation to obtain the 5th and 7th harmonic currents i on the d and q axes d5 (k+1), i d7 (k+1), i q5 (k+1), i q7 (k+1), 5th and 7th coordinate transformation formulas are as follows:

[0058]

[0059] According to the harmonic compensation unit structure diagram, the 5th and 7th harmonics filtered out by the d and q axes are integrated to obtain i dh 、i qh :

[0060] i dh =i d5 (k+1)+i d7 (k+1)

[0061] i qh =-i q5 (k+1)+i q7 (k+1)

[0062] The integrated harmonic current i dh 、i qh Cancel the original output current to get i′ d (k+1), i′ q (k+1), the calculation formula is as follows:

[0063] i′ d (k+1)=i d (k+1)-i dh

[0064] i′ q (k+1)=i q (k+1)-i qh .

[0065] Furthermore, the parameter k determines the bandwidth of the filter, and the value of k is 1.

[0066] Furthermore, the step (4) includes removing the harmonic current value i′ d (k+1), i′ q (k+1), is brought into the prediction model for modulation, and the cost function in step (1) is used to select the optimal voltage vector combination and allocate it to the inverter to drive the operation of the motor control system.

[0067] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: the present invention uses a frequency adaptive filter for harmonic compensation, which effectively reduces the adverse effects of harmonics generated by the nonlinearity and dead zone effect of the inverter on the prediction accuracy of the motor control system; it can perform adaptive adjustments according to the real-time changes in signal frequency, and can effectively improve the dynamic performance of the system; the software algorithm not only achieves high-performance filtering effects, but frequency adaptation also gives the system a high degree of flexibility, enabling it to easily cope with various complex application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1It is the structure diagram of the dual vector model predictive control system with current harmonic suppression;

[0069] Figure 2 It is the block diagram of the dual vector model predictive control system based on disturbance observer;

[0070] Figure 3 This is the structural block diagram of the harmonic compensation unit;

[0071] Figure 4 This is the structural block diagram of the harmonic current extraction unit. DETAILED DESCRIPTION

[0072] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0073] The present invention comprises the following steps:

[0074] (1) Based on the d-axis and q-axis voltage equations of the surface-mounted permanent magnet synchronous motor in the synchronous rotating coordinate system, a discretized current prediction model predicted by the dual-vector model is constructed;

[0075] The specific contents are as follows:

[0076] like Figure 1 The figure shows the structure of the dual vector model predictive control system with current harmonic suppression. The voltage equations of the permanent magnet synchronous motor on the d and q axes are as follows:

[0077]

[0078] Among them, u d with u q are the corresponding d-axis and q-axis voltages respectively; i d and i q is the d-axis and q-axis current; ω e is the electrical angular velocity; is the permanent magnet flux; R is the stator resistance. Since a surface-mounted permanent magnet synchronous motor is used, the inductances of the d and q axes are approximately equal.

[0079] The voltage equation can be rewritten as the current state equation as follows:

[0080]

[0081] Use the forward Euler discrete equation to predict the current i at the next moment d (k+1), i q (k+1), the specific current prediction equation is as follows:

[0082]

[0083] where i d (k), i q(k) represents the actual current value at the current moment, i d (k+1), i q (k+1) is the predicted current value at the next moment, u d (k),u q (k) represents the basic voltage vector of the inverter applied at the next moment. In the dual vector model predictive control, u d (k),u q (k) can be expressed as:

[0084]

[0085] The time t1 of the first vector action can be calculated according to the current zero-beat principle i d =0, the calculation formula is as follows:

[0086]

[0087] Wherein, S1 and S2 are the current change rates when the first vector and the second vector act on each other.

[0088] Using the mathematical model constructed above, the current generated by the composite voltage vector of two different vectors can be predicted. Using the cost function below, the optimal voltage vector combination with the smallest predicted current error can be selected to perform current prediction and regulation.

[0089]

[0090] (2) When parameter mismatch occurs, design the sliding surface and select the reaching law, and design the disturbance observer based on the sliding surface, reaching law and mathematical model;

[0091] like Figure 2 The figure shows the block diagram of the dual vector model predictive control system based on the disturbance observer. When the motor is actually running, there will be errors between the actual and predicted values ​​of the motor parameters. The d-axis and q-axis voltage equations on the stator side of the PMSM after adding the disturbance error can be expressed as follows:

[0092]

[0093] Among them, f d 、f q represents the disturbance error caused by the d-axis and q-axis parameters, F d 、F q is the parameter f d 、f q The rate of change.

[0094] Based on the above formula, the disturbance observer is designed as follows:

[0095]

[0096] in, denote the observed values ​​of the d-axis and q-axis currents, respectively, Respectively represent the observed values ​​of d-axis and q-axis disturbances; coefficient G d , G q represents the gain of sliding mode control; D dsmo , D qsmo Represents the sliding mode control function. The current error and disturbance error are defined as follows:

[0097]

[0098] Among them, e d 、e q is the error between the observed current value of d and q axis and the actual current value; e fd 、e fq is the error between the observed value of the d and q axis disturbance and the actual value of the disturbance. The observed value voltage equation and the actual value voltage equation are subtracted to obtain the error equation:

[0099]

[0100] Select e d 、e q As the sliding surface (ie: S d =e d , S q =e q ), the sliding mode control function is designed using the constant velocity reaching law, and the expression is as follows:

[0101]

[0102] The reaching law parameter k is a positive value.

[0103] The error equation can be rewritten as:

[0104]

[0105] Due to e fd , e fq is the disturbance quantity, and the sampling period is very short. It can be considered that the error change rate of the parameter disturbance is 0. Substituting it into the above formula, the sliding mode control function is obtained as follows:

[0106]

[0107] According to the discretized equation of current, a reasonable G d , G q The final disturbance observer mathematical model is as follows:

[0108]

[0109] (3) extracting the disturbance observer output current in a stationary coordinate system; extracting the fifth and seventh harmonics using a frequency adaptive filter controller, and transforming the harmonic current into a rotating coordinate system, performing feedback compensation with the original output current, and obtaining a current with the fifth and seventh harmonics removed;

[0110] The current equations containing the 5th and 7th harmonics in the d and q axes are as follows:

[0111]

[0112] To do this, it is necessary to filter out the 5th and 7th harmonic currents. Figure 3 As shown in the structural block diagram, the d and q axis currents output by the disturbance observer are transformed into the stationary coordinate system according to the inverse park transformation to obtain i α (k+1), i β (k+1), the transformation formula is as follows:

[0113]

[0114] Harmonic extraction is performed using a frequency adaptive filter. Figure 4 From the internal structure diagram of the harmonic extraction unit shown in the figure, we can know that the transfer function of the frequency adaptive filter is as follows:

[0115]

[0116] Where ω0 represents the resonant frequency of the filter. Based on the above formula, we can construct the input-output current transfer function based on the frequency adaptive filter:

[0117]

[0118] The parameter k determines the bandwidth of the filter. Taking into account both the dynamic performance and steady-state performance of the system, the value of k here is 1.

[0119] Inject 5ω into the frequency adaptive filter e 7ω e The resonant frequency is used to extract the 5th and 7th harmonic currents. The expressions are as follows:

[0120]

[0121] Extract the current containing the 5th and 7th harmonics in the stationary coordinate system: i α5 (k+1), i α7 (k+1), i β5 (k+1), i β7 (k+1).

[0122] Then perform park transformation to obtain the 5th and 7th harmonic currents i on the d and q axes.d5 (k+1), i d7 (k+1), i q5 (k+1), i q7 (k+1). The formulas for 5th and 7th order coordinate transformations are as follows:

[0123]

[0124] according to Figure 3 The harmonic compensation unit structure diagram shows that the 5th and 7th harmonics filtered out by the d and q axes are integrated to obtain i dh 、i qh :

[0125] i dh =i d5 (k+1)+i d7 (k+1)

[0126] i qh =-i q5 (k+1)+i q7 (k+1)

[0127] The integrated harmonic current i dh 、i qh Cancel the original output current to get i′ d (k+1), i′ q (k+1), the calculation formula is as follows:

[0128] i′ d (k+1)=i d (k+1)-i dh

[0129] i′ q (k+1)=i q (k+1)-i qh

[0130] (4) The current value i′ after removing the harmonics d (k+1), i′ q (k+1), substitute Figure 1 Modulation is performed in the prediction model shown, and the cost function in step 1 is used to select the optimal voltage vector combination and distribute it to the inverter to drive the operation of the motor control system.

Claims

1. A dual-vector model predictive control method with current harmonic suppression, characterized in that: The steps include: (1) Based on the d-axis and q-axis voltage equations of the surface-mounted permanent magnet synchronous motor in the synchronous rotating coordinate system, a discretized current prediction model predicted by the dual-vector model is constructed; (2) When parameter mismatch occurs, design the sliding surface and select the reaching law, and design the disturbance observer based on the sliding surface, reaching law and mathematical model; (3) extracting the disturbance observer output current in a stationary coordinate system; extracting the fifth and seventh harmonics using a frequency adaptive filter controller, and transforming the harmonic current into a rotating coordinate system, performing feedback compensation with the original output current, and obtaining a current with the fifth and seventh harmonics removed; (4) The current after harmonic filtering is injected into the prediction model for modulation, and the optimal voltage vector combination is selected to regulate the motor operation.

2. The dual-vector model predictive control method based on current harmonic suppression according to claim 1, characterized in that: The voltage equations of the surface-mounted permanent magnet synchronous motor on the d and q axes in step (1) are as follows: Among them, u d with u q are the corresponding d-axis and q-axis voltages respectively; i d and i q is the d-axis and q-axis current; ω e is the electrical angular velocity; is the permanent magnet flux; R is the stator resistance, and the inductances of the d and q axes are approximately equal.

3. The dual-vector model predictive control method based on current harmonic suppression according to claim 1, characterized in that: The process of constructing a discretized current prediction model predicted by a dual-vector model in step (1) includes: The voltage equation can be rewritten as the current state equation as follows: Use the forward Euler discrete equation to predict the current i at the next moment d (k+1), i q (k+1), the current prediction equation is as follows: Among them, i d (k), i q (k) represents the actual current value at the current moment, i d (k+1), i q (k+1) is the predicted current value at the next moment; u d (k) and u q (k) represents the basic voltage vector of the inverter applied at the next moment. In the dual vector model predictive control, u d (k) and u q (k) can be expressed as: The first vector action time t1, according to the current zero beat principle i d =0, the calculation formula is as follows: Wherein, S1 and S2 are the current change rates when the first vector and the second vector act on each other.

4. The dual-vector model predictive control method based on current harmonic suppression according to claim 3 is characterized in that: The discretized current prediction model is used to predict the current generated by the synthetic voltage vector of two different vectors, using the cost function The optimal voltage vector combination with the smallest predicted current error is selected to perform current prediction and regulation.

5. The dual-vector model predictive control method based on current harmonic suppression according to claim 1, characterized in that: The d-axis and q-axis voltage equations of the PMSM stator side with the disturbance error added in step (2) are expressed as follows: Among them, f d 、f q represents the disturbance error caused by the d-axis and q-axis parameters, F d 、F q is the parameter f d 、f q The rate of change.

6. The dual-vector model predictive control method based on current harmonic suppression according to claim 1, characterized in that: The disturbance observer is designed in step (2) as follows: in, denote the observed values ​​of d-axis and q-axis currents, respectively, Respectively represent the observed values ​​of d-axis and q-axis disturbances; coefficient G d , G q represents the gain of sliding mode control; D dsmo , D qsmo Represents the sliding mode control function, and defines the current error and disturbance error as follows: Among them, e d 、e q is the error between the observed current value of d and q axis and the actual current value; e fd 、e fq is the error between the observed value of the d and q axis disturbance and the actual value of the disturbance. The observed value voltage equation and the actual value voltage equation are subtracted to obtain the error equation: Select e d 、e q As the sliding surface, that is: S d =e d , S q =e q , the sliding mode control function is designed using the constant velocity reaching law, and the expression is as follows: The reaching law parameter k is a positive value, and the error equation is rewritten as: Due to e fd , e fq is the disturbance quantity, the sampling cycle time is very short, and the error change rate of the parameter disturbance is considered to be 0. Substituting it into the above formula, the sliding mode control function is: According to the discretized equation of current, a reasonable G d , G q Value, the disturbance observer mathematical model is as follows: in, Respectively represent the observed values ​​of d-axis and q-axis disturbances; coefficient G d , G q represents the gain of sliding mode control; i d (k), i q (k) represents the actual current value at the current moment.

7. The dual-vector model predictive control method based on current harmonic suppression according to claim 1, characterized in that: The current equations of the 5th and 7th harmonics in step (3) are as follows:

8. The dual-vector model predictive control method based on current harmonic suppression according to claim 1, characterized in that: The step (3) comprises: The d-axis and q-axis currents output by the disturbance observer are transformed into the stationary coordinate system according to the inverse Park transformation to obtain i α (k+1), i β (k+1), the transformation formula is as follows: Harmonic extraction is performed using a frequency adaptive filter. According to the internal structure diagram of the harmonic extraction unit, the transfer function of the frequency adaptive filter is as follows: Among them, ω0 represents the resonant frequency of the filter. According to the above formula, the input and output current transfer function based on the frequency adaptive filter is constructed: The parameter k determines the bandwidth of the filter; 5ω is injected into the frequency adaptive filter e , 7ω e The resonant frequency is used to extract the 5th and 7th harmonic currents. The expressions are as follows: Extract the current containing the 5th and 7th harmonics in the stationary coordinate system: i α5 (k+1), i α7 (k+1), i β5 (k+1), i β7 (k+1); Perform park transformation to obtain the 5th and 7th harmonic currents i on the d and q axes d5 (k+1), i d7 (k+1), i q5 (k+1), i q7 (k+1), 5th and 7th coordinate transformation formulas are as follows: According to the harmonic compensation unit structure diagram, the 5th and 7th harmonics filtered out by the d and q axes are integrated to obtain i dh 、i qh : i dh =i d5 (k+1)+i d7 (k+1) i qh =-i q5 (k+1)+i q7 (k+1) The integrated harmonic current i dh 、i qh Cancel the original output current to get i′ d (k+1), i′ q (k+1), the calculation formula is as follows: i′ d (k+1)=i d (k+1)-i dh i′ q (k+1)=i q (k+1)-i qh 。 9. The dual-vector model predictive control method based on current harmonic suppression according to claim 8, characterized in that: The parameter k determines the bandwidth of the filter, and the value of k is 1.

10. The dual-vector model predictive control method based on current harmonic suppression according to claim 1, characterized in that: The step (4) comprises: removing the harmonic current value i' d (k+1), i′ q (k+1), is brought into the prediction model for modulation, and the cost function in step (1) is used to select the optimal voltage vector combination and allocate it to the inverter to drive the operation of the motor control system.

Images