Error compensation method for dead-beat predictive current control model of high-speed permanent magnet synchronous motor in static coordinate system

Abstract

The invention discloses an error compensation method for a dead-beat predictive current control model of a high-speed permanent magnet synchronous motor in a static coordinate system. The method comprises the following steps: firstly, analyzing a dead-beat predictive current control scheme of the high-speed permanent magnet synchronous motor under a static coordinate system, and on the basis, analyzing a steady-state current control error caused by parameter mismatch, transient current tracking, voltage loss caused by nonlinearity of an inverter, and current errors in a steady state and a transient state; and an error compensation scheme based on dead-beat predictive current control of the high-speed permanent magnet synchronous motor under a static coordinate system is provided, namely, self-adaptive reference correction current injection is provided. According to the method, the parameter robustness of dead-beat predictive current control of the high-speed permanent magnet synchronous motor based on static coordinates is effectively improved, and steady-state current errors caused by inaccurate models can be effectively eliminated.

Description

Technical Field

[0001] The present invention relates to the technical field of motor drive control, and more particularly to an error compensation method for a deadbeat predictive current control model of a high-speed permanent magnet synchronous motor in a stationary coordinate system. Background Art

[0002] With the increasing demand for high-efficiency and high-power density high-speed permanent magnet synchronous motors (PMSMs) in many applications, especially in electrified transportation.

[0003] The drive of a high-speed permanent magnet synchronous motor usually has requirements of high electrical fundamental frequency and fast dynamic response. However, the switching frequency of power switches is limited, and the ratio of the switching frequency to the fundamental frequency (Switching-to-fundamental Frequency Ratio, SFR) becomes very low at high speeds. The control performance of traditional field-oriented control (FOC) deteriorates significantly at high speeds with a low SFR and may be even worse in the case of parameter mismatch. Therefore, advanced control technologies with fast dynamic response and strong parameter robustness are necessary for achieving high-performance high-speed permanent magnet synchronous motor drives.

[0004] Deadbeat predictive current control (DBPCC) has the advantages of fast dynamic response, the ability to reduce current harmonics at a constant switching frequency, and the combination of SVPWM, etc., which is more conducive to the high-bandwidth control of high-speed motors. At the same time, DBPCC can be divided into two categories, namely DBPCC based on a stationary coordinate system and DBPCC based on a synchronous rotating coordinate system, which depends on the coordinate system in which the reference voltage vector for implementing deadbeat control is initially derived. DBPCC based on a stationary coordinate system is more attractive for high-speed drives because it can more accurately compensate for control delay and rotor motion effects and can operate at a low SFR.

[0005] As a model-based control, motor parameters change due to manufacturing processes and operating conditions. Inaccurate motor parameters will lead to deterioration of control performance. For example, usually, an increase in temperature will cause a decrease in the PM flux linkage, and a high load current may reduce the motor inductance. However, the parameter robustness analysis and improvement research of DBPCC based on a synchronous rotating coordinate system are dominant, while the parameter robustness analysis and improvement research of DBPCC based on a stationary coordinate system are rarely reported. Therefore, improving the parameter robustness of DBPCC based on a stationary coordinate system is of great significance for achieving a high-performance and high-dynamic response high-speed PMSM system. Summary of the Invention

[0006] In view of the deficiencies in the prior art, the present invention provides an error compensation method for a deadbeat predictive current control model of a high-speed permanent magnet synchronous motor in a stationary coordinate system. It effectively improves the parameter robustness of the deadbeat predictive current control of a high-speed permanent magnet synchronous motor under a stationary coordinate system and can effectively eliminate the steady-state current error caused by inaccurate models.

[0007] To achieve the above object, the present invention adopts the following technical solutions:

[0008] In a first aspect, the present invention proposes an error compensation method for a deadbeat predictive current control model of a high-speed permanent magnet synchronous motor in a stationary coordinate system, and the method includes the following steps:

[0009] S1: Analyze the deadbeat predictive current control scheme of a high-speed permanent magnet synchronous motor in a stationary coordinate system.

[0010] S2: Analyze the steady-state current control error and transient current tracking error caused by parameter mismatch.

[0011] S3: Analyze the voltage loss caused by inverter non-linearity, and the current errors in steady state and transient state.

[0012] S4: According to the errors analyzed in S2 and S3, propose an error compensation scheme for the deadbeat predictive current control of a high-speed permanent magnet synchronous motor in a stationary coordinate system.

[0013] Further, in step S1, the deadbeat predictive current control scheme of a high-speed permanent magnet synchronous motor in a stationary coordinate system:

[0014] The deadbeat predictive current control includes the following steps:

[0015] The deadbeat predictive current control scheme of a high-speed permanent magnet synchronous motor in a stationary coordinate system takes into account control delay and the motion effect of the rotor. It mainly consists of three parts: calculating the reference stator flux through a current model, one-step prediction of the stator flux and current to compensate for one beat of inherent delay, and synthesis of the reference voltage to achieve deadbeat control. The current model is as follows:

[0016] ψ dq =L s i dq +ψ m

[0017] where ψ and i represent the stator flux vector and the stator current vector respectively, and the subscript dq represents the variables under the dq axis. L s represents the synchronous inductance of the surface-mounted permanent magnet synchronous motor, and ψ m is the permanent magnet flux. For convenience of representation, complex vectors are adopted in the present invention, and vectors are represented in bold.

[0018] Assume that the motor speed is constant within two control cycles. Then, the rotor position at time k + 1 is θ e (k)+ω e (k)T s , where θ e (k) and ω e (k) are the sampled rotor angular position and speed at time k respectively, and T s is the sampling period and also the control period. Based on this, the reference stator flux vector in the stationary coordinate system at time k + 2 can be calculated as:

[0019]

[0020] where the superscript * represents the reference value, the subscript αβ represents the variables under the α-β axes, e is the base of the natural logarithm, and e jθ means the vector rotates counterclockwise by θ degrees. Without considering the non-linearity of the inverter, the next stator flux vector can be predicted through the voltage model:

[0021] ψ αβ (k + 1) = ψ αβ (k)+T s u αβ (k) * -RT s i αβ (k)

[0022] where R is the phase resistance. Calculate the reference stator voltage vector u αβ (k) * , and use the current model and coordinate transformation at the current step to estimate the stator flux vector ψ αβ (k) at time k:

[0023]

[0024] Combining the above equation with the current model, the current at time k + 1 can be expressed as:

[0025]

[0026] Therefore, the reference stator voltage u αβ (k + 1) * can be finally synthesized according to the difference between the reference stator flux vector at time k + 2 and the estimated stator flux vector at time k + 1:

[0027]

[0028] It can be seen that the reference stator voltage vector required for deadbeat control is directly derived in the stationary coordinate system, avoiding the coordinate system conversion from the d-q axis to the α-β axis in the traditional DBPCC based on the synchronous coordinate system model. Therefore, this stationary coordinate system-based DBPCC is more suitable for high-speed permanent magnet synchronous motors with low SFR.

[0029] Furthermore, in step S2, the steady-state current control error and transient current tracking error caused by parameter mismatch are analyzed:

[0030] The parameter mismatch error analysis includes the following steps:

[0031] Since the resistance voltage drop is relatively small at high speeds, the resistance mismatch is ignored in the analysis. Assuming that the resistance voltage drop is negligible and using an ideal inverter, the stator flux linkage at time k + 2 is expressed as:

[0032]

[0033] where t k and t k+2 are the starting times, ψ αβ (k) is the stator flux linkage vector at time k, and u αβ is the actual instantaneous stator voltage vector. According to the principle of space vector modulation, its integral term can be expressed as:

[0034]

[0035] The reference voltage vector is calculated based on the estimated change in stator flux linkage:

[0036]

[0037] where ^ represents the estimated value, and are the stator flux linkages estimated using the estimated parameters at times k + 2, k + 1, and k, respectively. Combining the above equations gives:

[0038]

[0039] After Park coordinate transformation, we can obtain:

[0040]

[0041] Define the steady-state current error as Δi dq :

[0042] Δi dq = i dq * - i dq

[0043] According to the current model, it is assumed that the current error at time k+2 can be expressed as:

[0044]

[0045] Combining the above equations and assuming that the reference current at time k+2 is equal to the reference current at time k, the current error at time k+2 is:

[0046]

[0047] where ΔL s and Δψ m are the parameter mismatches of the inductor and the magnetic flux linkage, which are and respectively. In the steady state, if the current ripple caused by noise or other high-frequency interferences in the steady state is ignored, it can be assumed that the current values at times k+2 and k are equal:

[0048] i dq (k+2) = i dq (k)

[0049] Define I dq and as the steady-state current and the reference current respectively. Through the above analysis, the steady-state current value can be obtained as:

[0050]

[0051] Ideally, with accurate parameters, the current of this control method will reach the reference value within two time steps. Due to inaccurate parameters, it cannot track the reference value in the steady state, especially at high speeds. Therefore, the transient tracking performance under DBPCC parameter mismatch is related to the steady-state current rather than the actual reference current. The current error is rearranged as Δi dq_ss :

[0052]

[0053] At time k, define the current difference between the steady-state current I dq at this time and the current at time k+2 as the transient tracking error. Ideally, the tracking error is zero. In the case of inaccurate parameters, through the above equations, the tracking error err CT can be obtained as:

[0054]

[0055] It can be seen from the above equation that when the overshoot current or the transient tracking error appears two time steps after the change of the reference current and can quickly decrease within two time steps, so after a step change in the reference current, the tracking current error at 2n time steps is:

[0056]

[0057] Furthermore, in step S3, the voltage loss caused by inverter non-linearity, and the current errors in steady state and transient state:

[0058] Analysis steps for inverter non-linearity error:

[0059] Assume that compared with the dead-time effect, the parasitic voltage drop can be neglected. In the "i d = 0" control, the three-phase voltage losses caused by inverter non-linearity are V inv_a , V inv_b and V inv_c , which can be modeled as square waves in phase with the associated phase currents, with amplitudes of:

[0060] ΔV = V dc T dd / T s

[0061] By transforming the three-phase voltage losses to the d-q axis, the d-q axis voltage losses caused by inverter non-linearity can be derived. Under the "i d = 0" control, the average voltage loss on the d axis is zero, while the average voltage loss on the q axis is 4ΔV / π.

[0062] Generally, the average distorted voltage on the d-q axis caused by inverter non-linearity is considered. Therefore, the stator flux linkage equation for the ideal case at time k+2 in S2 is modified to:

[0063]

[0064] where V inv_dqav is the average voltage loss vector on the d-q axis, with the d-axis component being 0 and the q-axis component being 4ΔV / π. Through the same derivation as in S2, the steady-state current control error considering the effects of both parameter mismatch and inverter non-linearity can be obtained:

[0065]

[0066] It can be seen from the steady-state current error expression that the current control error caused by inverter non-linearity is similar to that caused by parameter mismatch and is independent of the reference current and the actual current. Therefore, the transient current tracking error of the distorted average voltage with inverter non-linearity can still be expressed in S2 and is only related to the parameter mismatch in the inductor. Therefore, the transient control performance of the proposed DBPCC will not be significantly affected by inverter non-linearity.

[0067] Furthermore, in step S4, an error compensation scheme for deadbeat predictive current control of a high-speed permanent magnet synchronous motor in a stationary coordinate system is proposed:

[0068] Steps of the error compensation scheme for deadbeat predictive current control:

[0069]

[0070] Among them, the coefficients λ L1 and λ L2 are related to inductance mismatch, λ ψ1 and λ ψ2 are related to flux linkage mismatch, λ inv1 and λ inv2 are related to inverter non-linearity, expressed as:

[0071]

[0072] The steady-state current control error compensation method proposed in the present invention is to inject two correction components into the original d-q axis reference currents to obtain new reference currents, and this error compensation method is called: Adaptive Reference Correction Current Injection. and are the original reference currents, COMP d and COMP q are the d-axis and q-axis reference correcting currents (Reference Correcting Currents, RCCs) respectively, and are the obtained new reference currents:

[0073]

[0074] Combining the above formula with the steady-state current, we can get:

[0075]

[0076] In order to eliminate the current control error, the required RCCs, and the corresponding coefficients can be solved as:

[0077]

[0078] C = -λ L1 λ ψ1 +λ L2 λ ψ2 -λ L1 λ inv1 +λ L2 λ inv2

[0079] D = -λ L2 λ ψ1 -λL1 λ ψ2 -λ L2 λ inv1 -λ L1 λ inv2

[0080] Injecting the desired RCCs into the original reference current will ultimately result in The equation is independent of both parameter mismatch and inverter nonlinearity. The current control error (average value) is ultimately zero. However, since the desired RCCs are related to the motor parameters, they cannot be directly obtained through calculation. Therefore, the present invention will adopt an adaptive scheme to derive the injected RCCs.

[0081] The objective function of the minimum current error can be listed, where e id and e iq are the sampled current errors on the d-axis and q-axis respectively:

[0082]

[0083] Substitute and into the equation of the injected correction current, and taking the difference can obtain:

[0084] e id =λ L1 ΔCOMP d +λ L2 ΔCOMP q

[0085] e iq =λ L1 ΔCOMP q -λ L2 ΔCOMP d

[0086]

[0087] Calculate the gradient vectors of the objective function J for COMP d and COMP q respectively, and then combine the above formula to obtain:

[0088]

[0089] According to the update principle of gradient descent, the amplitude of the injected RCCs is determined by the following formula:

[0090]

[0091] where η is the adaptive gain coefficient. Since λ L1It is a value that varies with the inductor parameters. When the parameters are accurate, it is equal to 1. When there is a parameter mismatch, it deviates from 1. In fact, when there is a parameter mismatch, it will be equal to a constant or change slowly and can be approximated as 1. The final adaptation law obtained is:

[0092] COMP d (k + 1)=COMP d (k)+ηe id (k)

[0093] COMP q (k + 1)=CoMP q (k)+ηe id (k)

[0094] In addition, due to the two-time-step delay of DBPCC, the current actual current should correspond to the reference value of two time steps. Therefore, the current error in the d-q axis is:

[0095]

[0096] Combining the above formula, it can be seen from the form of the formula that the amplitude of the injected RCCs can be realized by using the integrator with the d-q axis current error as the input:

[0097]

[0098] where K i is the integral gain, T s is the sampling duration, and the corresponding adaptive gain for minimizing the objective function J is:

[0099] η i =K i T s

[0100] 1. The present invention proposes an error compensation method for the deadbeat predictive current control model of a high-speed permanent magnet synchronous motor in the stationary coordinate system, which can avoid problems caused by coordinate transformation at high speeds due to rotor movement and is applicable to a wide speed range, including high speeds and low SFRs.

[0101] 2. The present invention proposes an error compensation method for the deadbeat predictive current control model of a high-speed permanent magnet synchronous motor in the stationary coordinate system, which avoids the step of calculating the correction voltage in the synchronous rotating coordinate system in the existing methods.

[0102] 3. The integral gain of the error compensation method for the deadbeat predictive current control model of a high-speed permanent magnet synchronous motor in the stationary coordinate system proposed by the present invention is exactly related to the elimination rate of the steady-state control error and is a simple design parameter, which enables the compensation of the steady-state error to be carried out quickly and is decoupled from the transient control at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0103] Figure 1 It is a flowchart of an error compensation method for a deadbeat predictive current control model of a high-speed permanent magnet synchronous motor in a stationary coordinate system according to an embodiment of the present invention.

[0104] Figure 2 It is a structural diagram of a deadbeat predictive current control method for a high-speed permanent magnet synchronous motor in a stationary coordinate system according to an embodiment of the present invention.

[0105] Figure 3 It is a structural diagram of an error compensation method for a deadbeat predictive current control model of a high-speed permanent magnet synchronous motor in a stationary coordinate system according to an embodiment of the present invention.

[0106] Figure 4 It is a schematic diagram of the speed effect of an error compensation method using a deadbeat predictive current control model of a high-speed permanent magnet synchronous motor in a stationary coordinate system according to an embodiment of the present invention.

[0107] Figure 5 It is a schematic diagram of the torque effect of an error compensation method using a deadbeat predictive current control model of a high-speed permanent magnet synchronous motor in a stationary coordinate system according to an embodiment of the present invention.

[0108] Figure 6 It is a schematic diagram of the d-q axis current effect before and after error compensation under 1.5 times inductance parameter mismatch according to an embodiment of the present invention.

[0109] Figure 7 It is a schematic diagram of the d-q axis current effect before and after error compensation under 1.5 times flux linkage parameter mismatch according to an embodiment of the present invention.

[0110] Figure 8 It is a schematic diagram of the d-q axis current effect before and after error compensation considering the inverter nonlinearity according to an embodiment of the present invention.

[0111] Figure 9 It is a schematic diagram of the d-q axis current effect before and after error compensation considering 0.9 times inductance parameter mismatch, 0.9 times flux linkage parameter mismatch and inverter nonlinearity simultaneously according to an embodiment of the present invention. Specific implementation scheme

[0112] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0113] Embodiment 1

[0114] Figure 1 It is a flowchart of an error compensation method based on a deadbeat predictive current control model of a high-speed permanent magnet synchronous motor in a stationary coordinate system. Figure 3It is the structure diagram of the error compensation method for the deadbeat predictive current control model of a high-speed permanent magnet synchronous motor based on a stationary coordinate system in an embodiment of the present invention. This embodiment proposes an error compensation method for the deadbeat predictive current control model of a high-speed permanent magnet synchronous motor under a stationary coordinate system, and the method includes the following steps:

[0115] S1: Analyze the deadbeat predictive current control method of a high-speed permanent magnet synchronous motor under a stationary coordinate system.

[0116] S2: Analyze the steady-state current control error and transient current tracking error caused by parameter mismatch.

[0117] S3: Analyze the voltage loss caused by inverter nonlinearity, and the current errors under steady state and transient state.

[0118] S4: Propose an error compensation method for the deadbeat predictive current control model of a high-speed permanent magnet synchronous motor based on a stationary coordinate system.

[0119] I. Deadbeat Predictive Current Control Method of High-Speed Permanent Magnet Synchronous Motor under Stationary Coordinate System

[0120] As Figure 2 shown, according to the reference d-q axis current, the current / magnetic flux model of the motor, and the coordinate transformation from the d-q axis to the α-β axis, predict the reference stator magnetic flux at the k+2 moment; according to the integral of the reference stator voltage vector at the k moment, predict the actual stator magnetic flux at the k+1 moment; synthesize the reference voltage at the k+1 moment and participate in the calculation; use the obtained voltage through the voltage limiting and space vector modulation module for PWM signal generation. Therefore, the reference d-q axis current can be tracked in two time steps, thereby obtaining a high control bandwidth.

[0121] II. Steady-State Current Control Error and Transient Current Tracking Error Caused by Parameter Mismatch

[0122] The steady-state current caused by parameter mismatch derived from the derivation is:

[0123]

[0124] When the motor parameters are inaccurate, the steady-state current will not be equal to the reference current, but a function of the reference current, parameter mismatch, true inductance, motor speed, and control time step. Based on the above, the following conclusion can be drawn: at very low speeds or sufficiently high sampling / control frequencies, one-step rotor movement ω e T s ≈0, that is This means that at low speeds and a sufficiently high sampling / control frequency, the steady-state current error due to parameter mismatch is negligible; for parameter mismatch only in the permanent magnet flux linkage, the current control error is independent of the reference current and is constant at a given speed, and the current control error increases with speed; the current control error caused by inductance parameter mismatch depends on the reference current and speed.

[0125] III. Voltage loss caused by inverter non-linearity, current error in steady state and transient state

[0126] The steady-state current considering parameter mismatch and inverter non-linearity derived from the derivation is:

[0127]

[0128] It can be seen that when the motor parameters are inaccurate and inverter non-linearity is considered, the steady-state current will not be equal to the reference current, but is a function of the reference current, parameter mismatch, true inductance, motor speed, control time step, and voltage distortion caused by dead-time effect. In addition to obtaining similar conclusions to those under parameter mismatch, it can also be seen that inverter non-linearity will cause a fixed offset error in the q-axis current, which is independent of the reference current. Moreover, the influence of inverter non-linearity is not significantly affected by speed and time step, and the q-axis component 2T associated with inverter non-linearity in the formula s V inv_dqav can be approximated as 4V dc T dd / π.

[0129] IV. Error compensation method for deadbeat predictive current control model of high-speed permanent magnet synchronous motor based on stationary coordinate system

[0130] In the embodiment of the present invention, error compensation of the deadbeat predictive current control model of the high-speed permanent magnet synchronous motor in the stationary coordinate system is used, and the parameter robustness of the deadbeat predictive current control of the high-speed permanent magnet synchronous motor based on the stationary coordinate is effectively improved by using the proposed method. In this example, the sampling frequency is set to be the same as the switching frequency. When the given initial speed is 1000 rpm, the given speed is 20000 rpm, and the working condition is to load 5 N·m at 0.2 s and 10 N·m at 0.3 s.

[0131] Figure 4 For the simulation speed waveform diagram, the SFR at this time is about 7.5, and a lower SFR can be achieved on the premise of instability. Figure 5 For the simulation torque waveform diagram, the torque response is rapid. Under the condition of ensuring that the given speed is consistent at 20000 rpm, a simulation comparison before and after compensation is carried out. Figure 6 For the schematic diagram of the d-q axis current effect before and after error compensation under 1.5 times inductance parameter mismatch; Figure 7Schematic diagram of d-q axis current effects before and after error compensation under 1.5 times flux linkage parameter mismatch; Figure 8 Schematic diagram of d-q axis current effects before and after error compensation considering inverter non-linearity; Figure 9 Schematic diagram of d-q axis current effects before and after error compensation considering 0.9 times inductance parameter mismatch, 0.9 times flux linkage parameter mismatch and inverter non-linearity. It can be seen that whether it is the influence of parameter mismatch or inverter non-linearity, the d-q axis current will have obvious deviation. After adopting the compensation method of adaptive reference correction current injection proposed by the present invention, the steady-state error of the d-q axis current is significantly reduced.

[0132] The method proposed by the present invention can well compensate the current control error caused by parameter mismatch and inverter non-linearity, and can maintain the robust transient performance of deadbeat predictive current control based on the stationary coordinate. This method is easy to implement and is also applicable to high-speed and low SFR situations.

Claims

1. A method for compensating an error of a deadbeat predictive current control model of a high-speed permanent magnet synchronous motor in a stationary coordinate system, comprising the following steps: S1: The deadbeat predictive current control scheme of high-speed permanent magnet synchronous motor in a stationary coordinate system is analyzed. S2: The steady-state current control error and transient current tracking error caused by parameter mismatch are analyzed. S3: The voltage loss caused by the nonlinearity of the inverter and the current error in steady state and transient state are analyzed. S4: According to the errors analyzed in S2 and S3, an error compensation scheme for zero-beat predictive current control of high-speed permanent magnet synchronous motor based on a stationary coordinate system is proposed.

2. The control method according to claim 1, characterized in that: In step S1, the deadbeat prediction current control scheme of the high-speed permanent magnet synchronous motor is based on the stationary coordinate system: The deadbeat predictive current control scheme for high-speed permanent magnet synchronous motors based on a stationary coordinate system takes into account control delays and rotor motion effects. It mainly consists of three parts: calculation of the reference stator flux through the current model, one-step prediction of the stator flux and current to compensate for one-beat inherent delay, and synthesis of the reference voltage to achieve deadbeat control. The current model is as follows: ψ dq =L s I dq +ψ m Where ψ and i represent the stator flux vector and stator current vector respectively, the subscript dq represents the variable under the dq axis, L s represents the synchronous inductance of the surface-mounted permanent magnet synchronous motor, ψ m is the permanent magnet flux. For the convenience of representation, complex vectors are used in the present invention, and vectors are represented in bold. Assuming that the motor speed is constant within two control cycles, the rotor position at time k+1 is θ e (k)+ω e (k)T s , where θ e (k) and ω e (k) are the sampled rotor angular position and speed at time k, T s is the sampling period and also the control period. Based on this, the reference stator flux vector in the stationary coordinate system at time k+2 can be calculated as: Wherein, the superscript * indicates the reference value, the subscript αβ indicates the variable under the α-β axis, e is the base of the natural logarithm, and e jθ Indicates that the vector rotates counterclockwise by θ degrees. Without considering the nonlinearity of the inverter, the next stator flux vector can be predicted by the voltage model: ψ αβ (k+1)=ψ αβ (k)+T s u αβ (k) * -RT s and αβ (k) Where R is the phase resistance. Calculate the reference stator voltage vector u αβ (k) * , and use it to calculate the current step. The stator flux vector ψ at time k αβ (k), the current model and coordinate transformation are estimated by: Combining the above formula and the current model, the current at time k+1 can be expressed as: Therefore, the reference stator voltage u at time k+1 is αβ (k+1) * The difference between the reference stator flux vector at time k+2 and the estimated stator flux vector at time k+1 can be finally synthesized: It can be seen that the reference stator voltage vector required for deadbeat control is directly derived in the stationary coordinate system. The related coordinate system conversion from dq axis to α-β axis in the conventional DBPCC based on synchronous coordinate system model is avoided. Therefore, this DBPCC based on stationary coordinate system is more suitable for high-speed permanent magnet synchronous motor with low SFR.

3. The control method according to claim 1, characterized in that: In step S2, the steady-state current control error and transient current tracking error caused by parameter mismatch are analyzed: Since the resistor voltage drop is relatively small at high speed, the resistor mismatch is ignored in the analysis. Assuming the resistor voltage drop is negligible and an ideal inverter is used, the stator flux at time k+2 is expressed as: where t k and t k+2 is the starting time, ψ αβ (k) is the stator flux vector at time k, u αβ is the actual instantaneous stator voltage vector. According to the principle of space vector modulation, its integral term can be expressed as: The reference voltage vector is calculated based on the estimated stator flux change: Where ^ represents an estimated value, and The stator flux estimated by using the estimated parameters at time k+2, k+1 and k respectively. Combining the above formulas, we get: After Parker coordinate transformation, we can get: Define the steady-state current error as Δi dq : Δi dq =i dq * -i dq According to the current model, it is assumed that the current error at time k+2 can be expressed as: Combining the above formula and assuming that the reference current at time k+2 is equal to the reference current at time k, the current error at time k+2 is: Where ΔL s With Δψ m is the parameter mismatch of inductance and flux linkage, respectively and In the steady state, if the current ripple caused by noise or other high-frequency interference in the steady state is ignored, it can be assumed that the current values ​​at time k+2 and time k are equal: i dq (k+2)=i dq (k) Definition I dq and They are steady-state current and reference current respectively. Through the above analysis, the steady-state current value can be obtained as follows: Ideally, with accurate parameters, the current of this control method will reach the reference value within two time steps. Due to inaccurate parameters, it is impossible to track the reference value in steady state, especially at high speed. Therefore, the transient tracking performance under DBPCC parameter mismatch is related to the steady state current, not the actual reference current. The current error is rearranged as Δi dq_ss : At time k, the steady-state current I is defined as dq The current difference at time k+2 and the transient tracking error. Ideally, the tracking error is zero. In the case of inaccurate parameters, the tracking error err can be obtained by the above formula CT : It can be seen from the above formula that when the overshoot current or transient tracking error occurs two time steps after the reference current changes, and can be quickly reduced within two time steps, the tracking current error of 2n time steps after the reference current has a step change is:

4. The control method according to claim 1, characterized in that: In step S3, the voltage loss caused by the nonlinearity of the inverter and the current error in steady state and transient state are: Assuming that the parasitic voltage drop is negligible compared to the dead-time effect, d = 0" control, the three-phase voltage loss caused by the nonlinearity of the inverter is V inv_a , V inv_b and V inv_c , they can be modeled as square waves in phase with the associated phase currents, with amplitudes of: ΔV=V dc T dd / T s By converting the three-phase voltage loss to the dq axis, the dq axis voltage loss caused by the inverter nonlinearity can be derived. d =0”, the average voltage loss on the d-axis is zero, while the average voltage loss on the q-axis is 4ΔV / π. Usually, the average distortion voltage of the dq axis caused by the nonlinearity of the inverter is considered. Therefore, the stator flux equation at time k+2 under ideal conditions is modified in S2 as follows: Where V inv_dqav is the average voltage loss vector of the dq axis, where the d-axis component is 0 and the q-axis component is 4ΔV / π. After repeated derivation with S2, the steady-state current control error considering the influence of both parameter mismatch and inverter nonlinearity can be obtained: From the steady-state current error expression, it can be seen that the current control error due to inverter nonlinearity is similar to the current control error caused by parameter mismatch and has nothing to do with the reference current and the actual current. Therefore, the transient current tracking error of the distorted average voltage with inverter nonlinearity can still be expressed in S2, which is only related to the parameter mismatch in the inductor. Therefore, the transient control performance of the proposed DBPCC will not be significantly affected by the inverter nonlinearity.

5. The control method according to claim 1, characterized in that: In step S4, an error compensation scheme based on the deadbeat predictive current control of the high-speed permanent magnet synchronous motor in a stationary coordinate system is proposed: The error compensation scheme steps of deadbeat predictive current control are as follows: The coefficient λ L1 and λ L2 Related to inductance mismatch, λ ψ1 and λ ψ2 Related to the flux mismatch, λ inv1 and λ inv2 It is related to the nonlinearity of the inverter and is expressed as: The steady-state current control error compensation method proposed by the present invention injects two correction components into the original dq-axis reference current to obtain a new reference current. The error compensation method is called: adaptive reference correction current injection. and is the original reference current, COMP d and COMP q are the d-axis and q-axis reference correcting currents (RCCs), and To obtain the new reference current: Combining the above equation with the steady-state current, we can get: In order to eliminate the current control error, the required RCCs and the corresponding coefficients can be solved as: C=-λ L1 l ψ1 +λ L2 l ψ2 -l L1 l inv1 +λ L2 l inv2 D=-λ L2 l ψ1 -l L1 l ψ2 -l L2 l inv1 -l L1 l inv2 Injecting the desired RCCs into the original reference current, we end up with The equation is independent of parameter mismatch and inverter nonlinearity, and the current control error (average value) is ultimately zero. However, since the desired RCCs are related to the parameters of the motor, they cannot be obtained directly by calculation. The present invention will adopt an adaptive scheme to derive the injected RCCs. The objective function of minimum current error can be listed, where e id and e iq They are the sampling current errors of the d-axis and q-axis respectively: Will and Substituting this into the equation for injecting the correction current, we get: e id =λ L1 ΔCOMP d +λ L2 ΔCOMP q e iq =λ L1 ΔCOMP q -l L2 ΔCOMP d COMP d and COMP q Calculate the gradient vector of the objective function J, and combine it with the above formula to get: According to the update principle of gradient descent, the amplitude of injected RCCs is determined by the following formula: Where η is the adaptive gain coefficient, due to λ L1 It is a value that changes with the inductor parameters. It is equal to 1 when the parameters are accurate and deviates from 1 when the parameters are mismatched. In fact, when the parameters are mismatched, it will be equal to a constant or change slowly, which can be approximated to 1. The final adaptation law is: COMP d (k+1)=COMP d (k)+ηe id (k) COMP q (k+1)=CoMP q (k)+ηe id (k) In addition, due to the two time step delay of DBPCC, the actual current should correspond to the reference value of two time steps, so the current error of dq axis is: Combining the above equations, it can be seen from the form of the equation that the amplitude of the injected RCCs can be realized by an integrator with the dq axis current error as input: Among them, K i is the integral gain, T s is the sampling duration, and the corresponding adaptive gain used to minimize the objective function J is: or i =K i T s 。

Images