Model-free deadbeat predictive current control method for permanent magnet synchronous motor based on current separation

Abstract

The present invention belongs to the technical field of motor drive and control applications, and proposes a model-free deadbeat predictive current control method for permanent magnet synchronous motors based on current separation. The main scheme is as follows: First, a current separation model of the permanent magnet synchronous motor is introduced, and the relationship between different parts in the sampled current change is derived based on this model; then, a model-free deadbeat control integral control law with extremely low parameter dependence is designed; further, within one control cycle, only a series of uncomplicated calculations are required, and more accurate predicted current and predicted voltage can be calculated without relying on motor parameters. Therefore, the present invention does not require any observer, avoiding the errors and oscillations that may be brought by the observer.

Description

Technical Field

[0001] The present invention relates to the technical field of motor drive and control applications, and particularly to a model-free deadbeat predictive current control method for permanent magnet synchronous motors based on current separation. Background Art

[0002] With the rapid development of permanent magnet material technology, permanent magnet materials are more and more applied in the industrial field. Due to characteristics such as high power density and high efficiency, permanent magnet synchronous motors (PMSMs) have been widely used in various fields such as aerospace, electric vehicles, industrial manufacturing, and agricultural production. Compared with traditional motors, PMSMs have the characteristics of simple structure, small rotor loss, and high power factor, making PMSMs the focus of research in recent years.

[0003] Field Oriented Control (FOC) and Direct Torque Control (DTC) are two classic methods in the PMSM control system. Although the above two methods already have good performance, there are still some problems in their control. In the actual PMSM control system, the key to determining its current and torque output capabilities is the current control performance. Since the control methods of traditional PMSM are not ideal in the case of sudden changes in motor speed and load changes, it is difficult to meet higher requirements. Therefore, in a complex control system with high performance, strong nonlinearity and strong coupling, advanced control methods are needed. Model Predictive Control (MPC), as an advanced intelligent control strategy, can handle nonlinear multi-objective parameters, has a fast dynamic response speed, and is suitable for multi-variable and strongly coupled PMSM systems. However, since the MPC control method depends on the accuracy of the motor model, and in actual working conditions, model parameters (such as stator resistance, magnetic flux, torque constant) are easily affected by factors such as temperature and magnetic saturation, resulting in a decline in control performance. Model-Free Predictive Control (MFPC) is a data-driven control method that avoids relying on an accurate motor model and can better adapt to parameter changes and complex nonlinear systems. Since the Deadbeat Predictive Current Control (DPCC) control strategy has better control performance and a simpler calculation process than the Model Predictive Current Control (MPCC), the Model-Free DPCC (MFDPCC) has been studied by many scholars. MFDPCC makes the current approach the reference current by estimating the optimal reference voltage output, and has excellent steady-state performance, but is more sensitive to changes in PMSM model parameters. Some scholars use a hyperlocal model and an error observer to estimate the motor parameter error, but in fact, the sensitivity to PMSM model parameters is transferred to the sensitivity to observer parameters, and the selection of observer parameters is very complicated.

[0004] The current separation method uses the proposed adaptive strategy to separate the natural response component and the forced response component in the current change amount, and at the same time calculates the current change amount corresponding to the unit voltage vector. By adaptively adjusting these two components, the actual current can easily track the reference current, avoiding the cumbersome parameter adjustment process of traditional MFDPCC. Some scholars use the current separation method to reduce the current prediction error, but the prediction error is still very large, and the control law parameters are selected as constants based on experience. Summary of the Invention

[0005] The object of the present invention is to provide a model-free deadbeat predictive current control method for permanent magnet synchronous motors based on current separation, aiming to solve the problems of parameter sensitivity and poor robustness existing in model predictive current control, and capable of estimating the parameters of the adaptive control law in each cycle, effectively reducing the dependence on the parameters of the control system and improving the parameter robustness of the system.

[0006] To solve its technical problems, the technical solution adopted by the present invention is as follows:

[0007] A model-free deadbeat predictive current control method for permanent magnet synchronous motors based on current separation includes the following steps:

[0008] Collect phase currents , , and bus voltage . Perform park transformation on the sampled phase currents to obtain dq stator currents and in the rotating coordinate system. At the same time, read the voltage vector k applied in the ( -1)th cycle, and then read the predicted voltage k calculated in the ( -1)th cycle and apply it to the inverter;

[0009] Subtract the feedback speed from the given speed , and obtain dq the reference current in the rotating coordinate system after calculation by a PI controller. At the same time, set d the axis reference current;

[0010] Take the difference between k sampled in the k th cycle and , sampled in the ( -1)th cycle to obtain the corresponding current change;

[0011] Read the natural response of the current k applied in the ( -1)th cycle, read the integral of the current prediction error k applied in the ( -1)th cycle, and read the unit voltage current change k applied in the (

[0012] k Read the k-1) Adaptive integral control law for cycle calculation to predict current , calculate the current prediction error per unit time , and calculate the integral of the current prediction error applied in the ( k ) cycle ;

[0013] Using the adaptive integral control law, calculate the adaptive integral coefficient in the ( k ) cycle ;

[0014] Set the thresholds of the adaptive integral coefficient and as the parameters for adaptive adjustment : Using the adaptive integral control law, calculate the natural response of the current applied in the (

[0015] ) cycle k ;

[0016] Calculate the change in current per unit voltage in the ( k ) cycle , thereby calculating the predicted current in the ( k + 1) cycle and the predicted voltage at the ( k + 1) moment , preparing for the inverter output in the next cycle;

[0017] Repeat the above steps in sequence to form a closed-loop control.

[0018] dq As a further optimization, the mathematical model of the permanent magnet synchronous motor in the rotating coordinate system is:

[0019] ,

[0020] where , dq are the stator currents in the axis rotating coordinate system respectively, , dq are the stator voltages in the L axis rotating coordinate system respectively, L , dq q is the inductance in the R , s and are the magnetic flux of the permanent magnet, the stator resistance and the electrical angular velocity of the rotor respectively.

[0021] dqAs a further optimization, the forward Euler method is used to discretize the mathematical model of the permanent magnet synchronous motor in dq the rotating coordinate system, which is expressed as:

[0022] ,

[0023] where T s is the control period, ([[]] k k ) and ([[]] k k + 1) correspond to the ([[]] k k )th and the ([[]] k k + 1)th control periods respectively.

[0024] As a further optimization, the current change is expressed as:

[0025] ,

[0026] In the formula, is the sampled current in the ([[]] k k - 1)th control period, is the sampled current in the ([[]] k k )th control period.

[0027] As a further optimization, the current prediction error per unit time is expressed as:

[0028] ,

[0029] where is the predicted current calculated in the ([[]] k k - 1)th control period;

[0030] The current prediction error integral k applied in the ([[]] )th period is expressed as:

[0031] ,

[0032] where is the current prediction error integral applied in the ([[]] k k - 1)th period.

[0033] As a further optimization, the adaptive integral control law is expressed as:

[0034] ,

[0035] where is the ([[]] k k)The adaptive integral coefficient of the periodic control law, is the upper threshold of is the lower threshold of

[0036] As a further optimization, the setting of the adaptive integral coefficient threshold and , as the parameters for adaptive adjustment means:

[0037] When , then adjust to ;

[0038] When , then adjust to ;

[0039] When , then the value of remains unchanged.

[0040] As a further optimization, the change in the unit voltage and current in the ( k )th cycle is expressed as:

[0041] ,

[0042] where is the predicted voltage calculated in the ( k - 1)th cycle;

[0043] The predicted current k in the ( + 1)th cycle k and the predicted voltage at the (

[0044] ,

[0045] where is the reference current.

[0046] The beneficial effects of the present invention are:

[0047] 1. The present invention directly obtains the current change information from the change in the sampled current, without relying on any motor model, improving the reliability of the control system.

[0048] 2. The adaptive control law proposed by the present invention can automatically update the integral parameters in each cycle, improving the accuracy of the predicted current in each cycle.

[0049] 3. The present invention only needs to set a threshold for the integral parameter of the control law within the stable region, with almost no parameter tuning process, greatly simplifying the complexity of the control system and improving the versatility of the control system.

[0050] 4. The present invention does not require any error observer, avoiding current oscillation and harmonic error that may be brought by the observer, and improving the stability of the control system. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is the principle block diagram of model-free deadbeat predictive current control in the embodiment of the present invention;

[0052] Figure 2 is the MFDPCC control flow chart in the embodiment of the present invention;

[0053] Figure 3 is the MFDPCC experimental waveform diagram in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and illustrated herein can be arranged and designed in various different configurations. Embodiment

[0055] Refer to Figure 1 the principle block diagram of model-free deadbeat predictive current control, which includes a speed outer loop PI regulator, a current prediction module, an inverter, a current sensor, and a permanent magnet synchronous motor. The model-free deadbeat predictive current control method for a permanent magnet synchronous motor based on current separation provided in this embodiment includes the following steps:

[0056] Step 1: Collect the phase currents , , and the bus voltage through the current sensor. Perform park transformation on the sampled phase currents to obtain the stator currents dq in the rotating coordinate system , . At the same time, read the voltage vector k applied in the ( -1)th period, and then read the predicted voltage k calculated in the ( -1)th period and apply it to the inverter.

[0057] Step 2: Subtract the feedback speed from the given speed ​ , which is obtained after calculation by the PI controller dq Reference current in the rotating coordinate system , and at the same time set d Axis reference current .

[0058] Step 3: Subtract the k -th period and the k ([-1])-th period sampled , to obtain the corresponding current change , specifically expressed as:

[0059] ,

[0060] In the formula, is the sampled current of the k ([-])-th control period, is the sampled current of the k ([-1])-th control period. This quantity can be obtained only through the sampled current without any motor model.

[0061] Step 4: Read the natural response of the current applied in the k ([-1])-th period , the integral of the current prediction error applied in the k ([-1])-th period , and the current change per unit voltage applied in the k ([-1])-th period . If the information read is 0, a smaller number, such as 0.01, is used to prevent calculation problems.

[0062] Step 5: Read the predicted current of the adaptive integral control law calculated in the k ([-1])-th period , calculate the current prediction error per unit time, and calculate the integral of the current prediction error k applied in the ， Current prediction error Specifically expressed as:

[0063] ,

[0064] Among them, is the predicted current calculated in the k ([-1])-th control period;

[0065] Integral of current prediction error Specifically expressed as:

[0066] ,

[0067] Among them, is the integral of the current prediction error applied in the ( k -1)th period, T s is the control period.

[0068] For the adaptive integral control law in this embodiment, the reason for its use is that in the actual experimental process, the parameter tuning process for the conventional control law or the model with an observer is relatively complex, and the current prediction error is greatly affected by these parameters. Taking the conventional control law as an example, even in the same motor, the design according to general parameters is usually a constant, and there will be a large difference in the magnitude of the predicted current error under different reference speed settings. Therefore, in this embodiment, the information of the sampled current can be directly used to estimate the parameters of the control law, greatly reducing the sensitivity of the control system to the parameters. At this time, using the adaptive parameters for model-free predictive control can obtain a more accurate predicted voltage and reduce the predicted current error.

[0069] Step six: Use the adaptive integral control law proposed in this embodiment to calculate the adaptive integral coefficient in the ( k )th period , and the specific derivation is as follows:

[0070] The mathematical model of the permanent magnet synchronous motor in the dq rotating coordinate system is:

[0071] ,

[0072] In the formula , are respectively the stator currents in the dq axis rotating coordinate system, , are respectively the stator voltages in the dq axis rotating coordinate system, L d , L q is dq the inductance in the , R s and are respectively the magnetic flux of the permanent magnet, the stator resistance and the electrical angular velocity of the rotor.

[0073] Discretize the above formula using the forward Euler method:

[0074] ,

[0075] Among them, T s is the control period, ( k ) and ( k +1) respectively correspond to the ( k )-th and the ( k +1)-th control periods.

[0076] According to the deadbeat control principle, the actual current tracks the reference current at the start of the next control period after the predicted voltage is applied to the inverter. Then, the predicted voltage can be expressed by the above formula as:

[0077] ,

[0078] Among them, 、 are dq axis reference currents.

[0079] Considering the one-step delay in the control system, the results calculated in the ( k )-th and the ( k +1)-th control periods will be loaded in the next control period. Then, one-step delay compensation is required:

[0080] ,

[0081] According to the current separation method, the change in the sampled current includes two parts:

[0082] ,

[0083] Among them, is the natural response to the current change and can be regarded as a constant in the steady state, is the forced response to the voltage change.

[0084] According to the PMSM discrete model, and can be expressed as:

[0085] ,

[0086] Among them, , , are all related to the PMSM model. In particular, can be expressed as:

[0087] ,

[0088] At this time, The meaning of can be regarded as the forced response of the current change to the unit voltage, denoted as the unit voltage current change. At this time, is redefined as :

[0089] ,

[0090] Among them, is the voltage vector applied to the inverter in the ( k )-th control period, that is, the predicted voltage calculated in the ( k -1)-th control period, which will cause a forced current response in the ( k )-th control period .

[0091] Therefore, the change in the sampled current of the current separation method can be expressed as:

[0092] ,

[0093] Furthermore, the predicted current in the next period can be expressed as:

[0094] ,

[0095] Considering one-step delay compensation, in order to make the sampled current in the next period approach the reference current, the above formula can be used to substitute the reference current for the predicted current to deduce the predicted voltage. The predicted reference voltage can be expressed as:

[0096] ,

[0097] It can be seen that the predicted reference voltage formula does not involve any content related to the PMSM parameters. At the same time, since the control period is short enough, there is no obvious change between two adjacent periods, so it can be considered that .

[0098] In summary, the basic control law of this control system is formed. Considering the robustness of the control system, the current prediction error integral information is added to the control law. Then the basic control law can be expressed as:

[0099] ,

[0100] Among them, is the current prediction error in the ( k )-th period, is the control law integral coefficient in the ( k )-th period, is the current prediction error integral calculated in the ( k -1)-th period.

[0101] In this embodiment, it is proved through the following process that can be directly expressed by the sampled current information:

[0102] The predicted reference voltage calculated in the ( k -1)th cycle can be expressed as:

[0103] ,

[0104] Thus, the natural response of the current change in the ( k )th cycle can be expressed as:

[0105] ,

[0106] Considering , from the above formula, we can obtain the expression:

[0107] ,

[0108] It can be seen that the integral coefficient of the control law can be obtained from the information of the current change and can be adaptively adjusted every cycle.

[0109] Step 7: Set the thresholds of the adaptive integral coefficient and as the parameters for adaptive adjustment of :

[0110] When , then adjust to ;

[0111] When , then adjust to ;

[0112] When , then the value of

[0113] In this embodiment, since the reference current is not stable during the speed change, it is necessary to set the thresholds of the adaptive integral coefficient and as the parameters for adaptive adjustment of to keep the control system stable. First, analyze the stability of the basic control law:

[0114] Through Z transformation, the transfer function of the basic control law is:

[0115] ,

[0116] List the characteristic equation of the calculation system, and we have:

[0117] ,

[0118] Thus, the characteristic roots can be calculated as:

[0119] ,

[0120] When the system is stable, all the characteristic roots must be inside the unit circle. It can be calculated that

[0121] ,

[0122] Get the value range of when the system is stable.

[0123] Next, as long as it is ensured that and , near the boundary, set the threshold of as and That's it. In this embodiment, adjust according to the following determination , which can keep the system stable:

[0124] When , then adjust to ;

[0125] When , then adjust to ;

[0126] When , then The value of remains unchanged.

[0127] Here, setting the threshold of as and is to make the adaptive parameter fall within the stable range of the control system. Assuming that the control law parameter is a constant, then the value range of the parameter within the stable domain of the system calculated from the transfer function of the control law is determined. However, since the reference current may change greatly when the speed is not stable, it may affect the calculation of the adaptive coefficient. Therefore, setting the threshold can prevent the adaptive coefficient from being too large or too small directly and making the system unstable. At the same time, in this embodiment, the designed threshold directly prevents the adaptive parameter from falling into the unstable domain, and the value of the threshold is very easy to determine, as long as it is near the boundary of the stable domain of the system. For the prediction error, it is automatically adjusted by the adaptive change of the coefficient.

[0128] Step Eight: Use the adaptive integral control law to calculate the natural response of the current applied in the ( k )th cycle , It can be expressed as:

[0129] ,

[0130] Step 9: Calculate the change in unit voltage and current in the ( k ) cycle, so as to calculate the predicted current in the ( k + 1) cycle and the predicted voltage at the moment of ( k + 1), preparing for the inverter output in the next cycle.

[0131] Specifically, it can be deduced by the current separation method :

[0132] ,

[0133] Furthermore, the predicted current can be expressed as:

[0134] ,

[0135] Thus, according to the calculation method mentioned before, the predicted voltage can be expressed as:

[0136] ,

[0137] Step 10: Repeat Steps 1 to 9 to form a closed-loop control. The above MFDPCC control process is as Figure 2 shown.

[0138] Finally, this embodiment is verified through experiments. The experimental data generates waveforms through MATLAB / Simulink. Figure 3 It is the experimental result under a reference speed of 600 rpm and a load torque of 5 Nm. When the system is stable, dq the shaft feedback current coincides with the reference current, indicating that the MFDPCC method proposed by the present invention has good current tracking performance.

[0139] The above is only the preferred embodiment of the present invention and is not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A model-free deadbeat predictive current control method for permanent magnet synchronous motors based on current separation, characterized in that Including the following steps: Collect phase current , , and bus voltage , perform park transformation on the sampled phase current to obtain dq stator current in the rotating coordinate system , . Meanwhile, read the voltage vector applied in the k -1 cycle, and then read the predicted voltage k calculated in the -1 cycle and apply it to the inverter; Subtract the feedback speed from the given speed , and obtain the reference current dq in the rotating coordinate system after calculation by the PI controller . At the same time, set d the axis reference current ; Subtract the samples obtained in the k th period from those in the k th period minus 1 to obtain the and . Take the difference to obtain the corresponding current change ; Read the natural response of the current applied in the k -1 cycle , read the integral of the current prediction error applied in the k -1 cycle , read the change in unit voltage current applied in the k -1 cycle ; Read the k adaptive integral control law prediction current for the -1 cycle calculation , calculate the current prediction error per unit time , and calculate the k integral of the current prediction error applied in the cycle ; Using the adaptive integral control law, calculate the k adaptive integral coefficient for the th period; Set the adaptive integration coefficient threshold and as the parameters for adaptive adjustment : Calculate the natural response of the current applied in the k th cycle using the adaptive integral control law ; Calculate the k periodic unit voltage and current changes , so as to calculate the k predicted current for the n+1th period k and the predicted voltage for the n+1th period, preparing for the inverter output in the next period; Repeat the above steps in sequence to form a closed-loop control.

2. The model-free deadbeat predictive current control method for a permanent magnet synchronous motor based on current separation according to claim 1, wherein The mathematical model of a permanent magnet synchronous motor in dq the rotating coordinate system is as follows: ， In the formula , are respectively dq the stator currents in the α-axis rotating coordinate system, , are respectively dq the stator voltages in the β-axis rotating coordinate system, L d , L q are dq the inductances in the α-axis rotating coordinate system, , R s and are respectively the magnetic flux linkage of the permanent magnet, the stator resistance and the electrical angular velocity of the rotor.

3. The model-free deadbeat predictive current control method for a permanent magnet synchronous motor based on current separation according to claim 2, wherein The mathematical model of the permanent magnet synchronous motor in the dq rotating coordinate system is discretized by the forward Euler method and expressed as: ， Among them, T s is the period, k and k +1 respectively correspond to the k and the k +(1) period.

4. The model-free deadbeat predictive current control method for permanent magnet synchronous motor based on current separation according to claim 3, characterized in that The current change is expressed as: ， Wherein, is the sampled current of the k -1 cycle, is the sampled current of the k cycle.

5. The model-free deadbeat predictive current control method for permanent magnet synchronous motors based on current separation according to claim 4, wherein The current prediction error within the unit time is expressed as: ， Among them, is the predicted current calculated for the k -1 cycle; The k integral of the predicted current error applied in the n-th cycle is expressed as: ， Among them, is the integral of the current prediction error applied in the k -1 cycle.

6. The model-free deadbeat predictive current control method for a permanent magnet synchronous motor based on current separation according to claim 4, characterized in that The adaptive integral control law is expressed as: ， Among them, is the control law adaptive integral coefficient of the k cycle, is the upper threshold of and is the lower threshold of 7. The model-free deadbeat predictive current control method for permanent magnet synchronous motors based on current separation according to claim 1, wherein The set adaptive integration coefficient threshold and , as parameters for adaptive adjustment , refers to: When then adjust to ; When then adjust to ; When then the value remains unchanged.

8. The model-free deadbeat predictive current control method for a permanent magnet synchronous motor based on current separation according to claim 1, wherein The k periodic unit voltage and current variation is expressed as: ， Among them, is the predicted voltage calculated for the k -1 cycle; The k +1 cycle predicted current and k +1 cycle predicted voltage are expressed as: ， Among them, is the reference current.

Images