An Improved Three-Level Model-Free Predictive Current Control Method for Permanent Magnet Synchronous Motor

Abstract

The improved three-level model-free predictive current control method of permanent magnet synchronous motor of the present invention first determines the relevant current reference value by giving the rotation speed; then judges the type of the currently applied voltage vector and the voltage set number, and then calculates the current difference corresponding to the basic voltage in each voltage set based on the current dq axis current difference through the corresponding rules, thereby realizing the calculation of the dq axis current difference used by all voltage vectors and updating the current difference lookup table; then calculates the predicted value of the dq axis current at the k+1 moment, and outputs the basic voltage vector that minimizes the value function through rolling optimization. The present invention fully adapts to the market trend of three-level inverter application, and the corresponding control does not require specific motor parameters to participate in the calculation, effectively alleviates the influence of model adaptation on system control performance, and effectively solves the problem of stagnation of current difference update in traditional model-free predictive current control.

Description

Technical Field

[0001] The invention relates to an improved three-level model-free predictive current control method for a permanent magnet synchronous motor, and belongs to the field of motor drive and control. Background Art

[0002] Nowadays, three-level inverters represented by three-level neutral point clamped (3L-NPC) inverters have attracted widespread attention in the field of high-power electronics and motor drives. Compared with traditional two-level inverters, three-level inverters have certain advantages in alleviating voltage distortion, reducing semiconductor stress and switching frequency. Therefore, this topology has become an attractive research topic in industry and academia. At the same time, in practical applications, motor parameters are not only difficult to accurately obtain, but also change with the changes in the operating environment. The model mismatch problem cannot be avoided, which greatly limits the scope of application of MPCC. Based on this, in order to reduce the impact of parameter uncertainty on system control performance, a model-free predictive current control (MFPCC) algorithm based on current difference detection has been proposed.

[0003] The MFPCC algorithm replaces the model-based current prediction by using the current difference under different switch states stored in the current difference lookup table in the past and the current state, without requiring any motor parameters to participate in the calculation. However, it has the problem of stagnation in updating the current difference lookup table that cannot be ignored. The existing forced update method is to force the output of the voltage vector as the final vector of the drive motor when it is detected that one (or more) voltage vectors have not been applied in several consecutive sampling cycles, but at the same time sacrifices the optimal output of the value function. Summary of the invention

[0004] Technical problem: To address the above problems, an improved three-level model-free predictive current control method for permanent magnet synchronous motor is proposed to achieve full update of the current difference within one sampling period, which can effectively increase the update frequency of the current difference and reduce the computational burden of the system by setting vector candidate rules.

[0005] Technical solution: An improved three-level model-free predictive current control method for a permanent magnet synchronous motor, characterized in that it comprises the following steps:

[0006] Step 1: Obtain the electrical angle θ of the permanent magnet synchronous motor through the encoder, and obtain the three-phase stator current i at time (k-1) and time k through the current sensor s (k-1) and i s (k), s = a, b, c, and then the three-phase stator current is transformed by Clark and Park to obtain the components i of the stator current on the dq axis at time (k-1) and time kd (k-1), i q (k-1) and i d (k), i q (k);

[0007] Step 2: Determine the voltage vector V applied at time (k-1) k The type and voltage set number of the voltage set are used to calculate the dq axis current difference Δi corresponding to the basic voltage vector in each voltage set. d |V i and Δi q |V i , and then complete the calculation of the dq axis current difference corresponding to all voltage vectors according to the relationship between the current differences in each voltage set, and realize the update of the current difference lookup table;

[0008] Step 3: Use the current prediction module and the current difference lookup table to predict the dq axis currents under different switching states at time (k+1) to obtain the predicted value i d (k+1)|V j and i q (k+1)|V j ;

[0009] Step 4: After the candidate vector is set, obtain different voltage vectors V j The value function output g under j ={g1,g2,…,g 27}; Then, the basic voltage vector that minimizes the cost function is output through rolling optimization, and finally, the optimal switching state is output through the midpoint potential balancing module to drive the inverter.

[0010] Furthermore, in step 2, the dq axis current difference Δi corresponding to the basic voltage vector in each voltage set d |V i and Δi q |V i The calculation method is as follows: First, determine the voltage vector V applied at time (k-1) according to Table 1 k The voltage collection number and its vector type are determined:

[0011] Table 1

[0012] Category Voltage vectors 0 <![CDATA[V 25 ,V 26 ,V 27 ]]> 1 <![CDATA[V1,V2,V3,V 13 ,V 14 ,V 15 ]]> 2 <![CDATA[V4,V 16 ]]> 3 <![CDATA[V5,V6,V7,V 17 ,V 18 ,V 19 ]]> 4 <![CDATA[V8,V 20 ]]> 5 <![CDATA[V9,V 10 ,V 11 ,V 21 ,V 22 ,V 23 ]]> 6 <![CDATA[V 12 ,V 24 ]]>

[0013] If V k Belongs to the zero vector, then Δi d |V 25、26、27 =δi d0 =Δi d |S k-1 , Δi q |V 25、26、27=δi q0 =Δi q |S k-1 ; Among them, δi d0 ,δi q0 Represents the natural attenuation of the dq axis current; Δi d |S k-1 and Δi q |S k-1 represents the dq axis current difference under the voltage vector at time (k-1),

[0014] If V k If it is a small vector, then the dq axis current difference Δi of the basic voltage vector in the set of all voltage vectors is calculated according to formula (1): d |V i and Δi q |V i , where V i Specifically, V1, V4, V5, V8, V9, V 12 、V 25 ;

[0015]

[0016] Among them, β di , β qi Represents the voltage vector V i The angle between the d-axis and the q-axis; β dk , β qk Represents the voltage vector V k The angle between the d-axis and the q-axis; δi d |V k ,δi q |V k Represents the voltage vector V k The forced response of the dq axis current under the action of The parameter values ​​are shown in Table 2;

[0017] Table 2

[0018]

[0019] In the table, V x Represents the amplitude coefficient between characteristic voltage vectors;

[0020] If V k If it is a large vector, the basic voltage vector V in its voltage set is calculated according to formula (2): m The corresponding dq axis current difference Δi d |V m and Δi q |V mThen, according to formula (3), the dq axis current difference Δi corresponding to the basic voltage vector in the set of all voltage vectors is calculated: d |V i and Δi q |V i ;

[0021]

[0022]

[0023] Among them, δi d |V m and δi q |V m They are voltage vector V m The forced response of dq axis current under the action;

[0024] If V k Belongs to the middle vector and is the basic vector, then directly calculate the dq axis current difference Δi corresponding to the basic voltage vector in the collection of all voltage vectors according to formula (1): d |V i and Δi q |V i ; If V k If it belongs to the middle vector but not to the basic vector, the basic voltage vector V in the voltage set to which it belongs is calculated according to formula (4): n The corresponding dq axis current difference Δi d |V n and Δi q |V n Then, according to formula (5), the dq axis current difference Δi corresponding to the basic voltage vector in the set of all voltage vectors is calculated: d |V i and Δi q |V i ;

[0025]

[0026]

[0027] Among them, δi d |V n and δi q |V n They are voltage vector V n The forced response of the dq axis current under the action of.

[0028] Further, in step 2, the method for implementing the updating of the current difference lookup table is: after obtaining the dq axis current difference corresponding to the basic voltage vector of the voltage set 1, 3, and 5, the dq axis current difference corresponding to other voltage vectors in the same set is calculated by formula (6); the dq axis current difference corresponding to all other voltage vectors of the voltage set 2, 4, and 6 is calculated by formula (7); the calculated dq axis current difference under different switching states is stored in a current difference lookup table containing 27 different switching states and the original data of the same switching state in the table is replaced to complete the updating of the current difference lookup table;

[0029]

[0030]

[0031] Among them, parameter a takes 1, 3, 5, and parameter b takes 2, 4, 6, corresponding to the voltage vector number.

[0032] Furthermore, the step 3 specifically includes: firstly obtaining the dq axis current difference Δi under different switching states from the current difference lookup table d |V j and Δiq|V j ; Then, according to formula (8), the predicted current value i of the dq axis under different switching states at time (k+1) is calculated d (k+1)|V j and i q (k+1)|V j ;

[0033]

[0034] Furthermore, in step 4, the vector candidate rule is: consider the basic vector in the adjacent sector of the action vector at the previous sampling moment as the candidate vector, take only one phase switch state continuously jumps in one sampling cycle as the second principle, screen small vectors, medium vectors, and large vectors, and then complete the small, medium, large, and zero vector types of the screened vectors, and finally the number of candidate vectors is 6-9;

[0035] Table 3

[0036]

[0037]

[0038] Table 3 shows the final candidate rules.

[0039] Beneficial effects: Based on a permanent magnet synchronous motor powered by a three-level inverter, the present invention constructs a current prediction model based on a current difference lookup table, avoids the participation of motor parameters in calculations, and improves the robustness of parameters. This method realizes full update of current difference within one cycle, effectively increasing the update frequency of current difference. At the same time, the vector candidate method adopted reduces the candidate vectors to 6-9, reducing the calculation burden of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A schematic diagram of the three-level model-free predictive current control of a permanent magnet synchronous motor provided by the present invention;

[0041] Figure 2 This is a simulation diagram of the q-axis current tracking performance of the MFPCC algorithm proposed in the present invention. Figure 3 (a) is the current tracking simulation under the condition of accurate parameters (R = 5.25 ohms). Figure 3 (b) is the current tracking simulation under the condition of 50% increase in resistance (R = 7.875 ohms);

[0042] Figure 3 Shown are the three-phase current waveforms of the MFPCC algorithm proposed in the present invention. DETAILED DESCRIPTION

[0043] The present invention will be further described in detail below with reference to the accompanying drawings and by way of examples. The following examples are intended to explain the present invention but the present invention is not limited to the following examples.

[0044] An improved three-level model-free predictive current control principle diagram of permanent magnet synchronous motor is shown in the figure. Figure 1 As shown, it includes a speed outer loop PI controller module 1, a minimization value function module 2, a midpoint potential balance module 3, an inverter module 4, a permanent magnet synchronous motor module 5, an encoder module 6, a current difference calculation module 7, a current difference lookup table module 8 and a current prediction module 9.

[0045] like Figure 2 As shown, the method comprises the following steps:

[0046] Step 1: Obtain the reference q-axis current i at time (k+1) according to the speed outer loop PI controller q ref (k+1):

[0047] The given speed N r ref With the actual speed N r The difference e n The current is sent to the speed outer loop PI controller, and the reference q-axis current i at time (k+1) is obtained according to formula (1): q ref (k+1):

[0048]

[0049] Among them, k p and k i are the proportional gain and integral gain of the speed PI controller respectively, and s is a complex variable.

[0050] Step 2: Obtain the electrical angle θ of the permanent magnet synchronous motor from the encoder; then use the current sensor to measure the three-phase stator current i of the permanent magnet synchronous motor at time (k-1) and time k respectively. s (k-1) and i s (k), s = a, b, c, after Clark transformation of formula (2), the αβ axis components i of the stator current at time (k-1) and time k are obtained α (k-1), i β (k-1) and i α (k), i β (k), and then after Park transformation of formula (3), the dq axis component i of the stator current at time (k-1) and time k is obtained d (k-1), i q (k-1) and i d (k), i q (k);

[0051]

[0052]

[0053] Step 3: Determine the voltage vector V applied at time (k-1) according to Table 1 k The voltage collection number and its vector type are determined:

[0054] Table 1

[0055] Category Voltage vectors 0 <![CDATA[V 25 ,V 26 ,V 27 ]]> 1 <![CDATA[V1,V2,V3,V 13 ,V 14 ,V 15 ]]> 2 <![CDATA[V4,V 16 ]]> 3 <![CDATA[V5,V6,V7,V 17 ,V 18 ,V 19 ]]> 4 <![CDATA[V8,V 20 ]]> 5 <![CDATA[V9,V 10 ,V 11 ,V 21 ,V 22 ,V 23 ]]> 6 <![CDATA[V 12 ,V 24 ]]>

[0056] If V k Belongs to the zero vector, then the corresponding dq axis current differences are Δi d |V 25、26、27 =δi d0 =Δi d |S k-1 , Δi q |V 25、26、27 =δi q0 =Δi q |S k-1 ; Among them, δi d0 ,δi q0 Represents the natural attenuation of the dq axis current; Δi d |S k-1and Δi q |S k-1 represents the dq axis current difference under the voltage vector at time (k-1),

[0057] If V k If it is a small vector, then the dq axis current difference Δi of the basic voltage vector in the set of all voltage vectors is calculated according to formula (4): d |V i and Δi q |V i , where V i Specifically, V1, V4, V5, V8, V9, V 12 、V 25 ;

[0058]

[0059] Among them, β di , β qi Represents the voltage vector V i The angle between the d-axis and the q-axis; β dk , β qk Represents the voltage vector V k The angle between the d-axis and the q-axis; V in the formula i 、V k Respectively represent vector V i With V k The amplitude coefficient, δi d |V k ,δi q |V k Represents the voltage vector V k The forced response of the dq axis current under the action of The parameter values ​​are shown in Table 2;

[0060] Table 2

[0061]

[0062] In the table, V x Represents the amplitude coefficient between characteristic voltage vectors;

[0063] If V k If it is a large vector, the dq axis current difference Δi corresponding to the basic voltage vector in the voltage set is calculated according to formula (5): d |V m and Δi q |V m Then, according to formula (6), the dq axis current difference Δi corresponding to the basic voltage vector in the set of all voltage vectors is calculated: d |V i and Δi q|V i ;

[0064]

[0065]

[0066] Among them, δi d |V m and δi q |V m They are voltage vector V m The forced response of dq axis current under the action;

[0067] If V k Belongs to the middle vector and is the basic vector, then directly calculate the dq axis current difference Δi corresponding to the basic voltage vector in the collection of all voltage vectors according to formula (4): d |V i and Δi q |V i ; If V k Belongs to the middle vector, but not to the basic vector, then first calculate the basic voltage vector V in the voltage set where it is located according to formula (7): n The corresponding dq axis current difference Δi d |V n and Δi q |V n , and then calculate the dq axis current difference Δi corresponding to the basic voltage vector in the set of all voltage vectors according to formula (8): d |V i and Δi q |V i ;

[0068]

[0069]

[0070] Among them, δi d |V n and δi q |V n They are voltage vector V n The forced response of the dq axis current under the action of.

[0071] Step 4: After obtaining the dq-axis current difference corresponding to the basic voltage vector of the voltage set 1, 3, and 5, the dq-axis current difference corresponding to other voltage vectors in the same set is calculated by formula (9); the dq-axis current difference corresponding to all other voltage vectors of the voltage set 2, 4, and 6 is calculated by formula (10); the calculated dq-axis current difference under different switching states is stored in a current difference lookup table containing 27 different switching states and the original data of the same switching state in the table is replaced to complete the update of the current difference lookup table.

[0072]

[0073]

[0074] Among them, parameter a takes 1, 3, 5, and parameter b takes 2, 4, 6, corresponding to the voltage vector number.

[0075] Step 5: Obtain the dq axis current difference Δi under different switching states from the current difference lookup table d |V j and Δi d |V j ; Then, according to formula (11), the predicted current value i of the dq axis under different switching states at time (k+1) is calculated d (k+1)|V j and i q (k+1)|V j .

[0076]

[0077] Step 6: Consider the basic vectors in the adjacent sectors of the action vector at the previous sampling moment as candidate vectors, and use the second principle that only one phase switch state changes continuously in one sampling cycle to screen small vectors, medium vectors, and large vectors. To improve the accuracy of the optimal vector selection, the small, medium, large, and zero vector types of the screened vectors are supplemented. The final number of candidate vectors is 6-9, as shown in Table 3.

[0078] Vector Pre-selection Vector Pre-selection <![CDATA[V1]]> <![CDATA[V1 V3 V4 V5 V 21 V 24 V 26 ]]> <![CDATA[V 15 ]]> <![CDATA[V 11 V 12 V 14 V 15 V 16 V 19 V 26 ]]> <![CDATA[V2]]> <![CDATA[V2 V3 V4 V6 V7 V 22 V 23 V 24 V 26 ]]> <![CDATA[V 16 ]]> <![CDATA[V 12 V 13 V 15 V 16 V 18 V 19 V 20 V 26 ]]> <![CDATA[V3]]> <![CDATA[V1 V3 V4 V7 V 23 V 24 V 26 ]]> <![CDATA[V 17 ]]> <![CDATA[V 13 V 16 V 17 V 19 V 20 V 21 V 26 ]]> <![CDATA[V4]]> <![CDATA[V1 V3 V4 V5 V7 V8 V 24 In 26 ]]> <![CDATA[V 18 ]]> <![CDATA[V 14 V 15 V 16 V 18 V 19 V 20 V 22 V 23 V 26 ]]> <![CDATA[V5]]> <![CDATA[V1 V3 V4 V5 V7 V8 V9 V 11 In 26 ]]> <![CDATA[V 19 ]]> <![CDATA[V 15 V 16 V 17 V 19 V 20 V 23 V 26 ]]> <![CDATA[V6]]> <![CDATA[V2 V4 V6 V7 V8 V 10 In 26 ]]> <![CDATA[V 20 ]]> <![CDATA[V 16 V 18 V 19 V 20 V 21 V 23 V 24 V 26 ]]> <![CDATA[V7]]> <![CDATA[V3 V4 V6 V7 V8 V 11 In 26 ]]> <![CDATA[V 21 ]]> <![CDATA[V1 V3 V 17 V 19 V 20 V 21 V 23 V 24 V 26 ]]> <![CDATA[V8]]> <![CDATA[V4 V5 V7 V8 V 10 V 11 V 12 V 26 ]]> <![CDATA[V 22 ]]> <![CDATA[V2 V 18 V 20 V 22 V 23 V 24 V 26 ]]> <![CDATA[V9]]> <![CDATA[V5 V8 V9 V 11 V 12 V 13 V 26 ]]> <![CDATA[V 23 ]]> <![CDATA[V3 V 19 V 20 V 22 V 23 V 24 V 26 ]]> <![CDATA[V 10 ]]> <![CDATA[V6 V7 V8 V 10 V 11 V 12 V 14 V 15 V 26 ]]> <![CDATA[V 24 ]]> <![CDATA[V2 V3 V4 V 20 V 21 V 23 V 24 V 26 ]]> <![CDATA[V 11 ]]> <![CDATA[V7 V8 V9 V 11 V 12 V 15 V 26 ]]> <![CDATA[V 25 ]]> <![CDATA[V1 V3 V9 V 17 In 25 ]]> <![CDATA[V 12 ]]> <![CDATA[V8 V 10 V 11 V 12 V 13 V 15 V 16 V 26 ]]> <![CDATA[V 26 ]]> <![CDATA[V2 V4 V5 V 10 V 13 V 18 V 21 V 26 ]]> <![CDATA[V 13 ]]> <![CDATA[V9 V 11 V 12 V 13 V 15 V 16 V 17 V 19 V 26 ]]> <![CDATA[V 27 ]]> <![CDATA[V6 V 14 V 15 V 22 V 25 ]]> <![CDATA[V 14 ]]> <![CDATA[V 10 V 12 V 14 V 15 V 16 V 18 V 26 ]]>

[0079] The method of the present invention first obtains the three-phase stator current i at time (k-1) and time k. s (k-1) and i s (k), s = a, b, c, rotor electrical angle θ; then the reference q-axis current i at time (k+1) is obtained through the PI controller q ref (k+1) and given the d-axis current reference i d ref (k+1)=0, and the components i of the stator current on the dq axis at time (k-1) and time k are obtained through step 2.d (k-1), i q (k-1) and i d (k), i q (k); perform step 3 to calculate the dq axis current difference corresponding to the basic voltage of each voltage vector set; then perform step 4 to update the current difference lookup table; then perform step 5 and the current difference lookup table updated in step 4 to calculate the dq axis current prediction value at time k+1, and perform rolling optimization in step 6 to output the basic voltage vector u that minimizes the value function min ;

[0080] The simulation results of an improved three-level model-free predictive current control for permanent magnet synchronous motor are shown in Figure 2. Figure 2 , Figure 3 As shown. Figure 2 (a) shows that under the condition of accurate motor parameters, the actual current can track the reference current very well, while Figure 2 As can be seen from (b), the MFPCC method proposed in the present invention does not require any motor parameters to participate in the calculation, and therefore can always maintain good q-axis current tracking performance. Figure 3 The three-phase current waveform is shown, and it can be seen that the positive current limit is good.

[0081] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An improved three-level model-free predictive current control method for a permanent magnet synchronous motor, characterized in that: The steps include: Step 1: Obtain the electrical angle θ of the permanent magnet synchronous motor through the encoder, and obtain the three-phase stator current i at time (k-1) and time k through the current sensor s (k-1) and i s (k), s = a, b, c, and then the three-phase stator current is transformed by Clark and Park to obtain the components i of the stator current on the dq axis at time (k-1) and time k d (k-1), i q (k-1) and i d (k), i q (k); Step 2: Determine the voltage vector V applied at time (k-1) k The type and voltage set number of the voltage set are used to calculate the dq axis current difference Δi corresponding to the basic voltage vector in each voltage set. d |V i and Δi q |V i , and then complete the calculation of the dq axis current difference corresponding to all voltage vectors according to the relationship between the current differences in each voltage set, and realize the update of the current difference lookup table; Step 3: Use the current prediction module and the current difference lookup table to predict the dq axis currents under different switching states at time (k+1) to obtain the predicted value i d (k+1)|V j and i q (k+1)|V j ; Step 4: After the candidate vector is set, obtain different voltage vectors V j The value function output g under j ={g1,g2,…,g 27 }; Then, the basic voltage vector that minimizes the value function is output through rolling optimization, and finally, the optimal switching state is output through the midpoint potential balancing module to drive the inverter; In step 2, the dq axis current difference Δi corresponding to the basic voltage vector in each voltage set d |V i and Δi q |V i The calculation method is as follows: First, determine the voltage vector V applied at time (k-1) according to Table 1 k The voltage collection number and its vector type are determined: Table 1 If V k Belongs to the zero vector, then Δi d |V 25、26、27 =δi d0 =Δi d |S k-1 , Δi q |V 25、26、27 =δi q0 =Δi q |S k-1 ; Among them, δi d0 ,δi q0 Represents the natural attenuation of the dq axis current; Δi d |S k-1 and Δi q |S k-1 represents the dq axis current difference under the voltage vector at time (k-1), If V k If it is a small vector, then the dq axis current difference Δi of the basic voltage vector in the set of all voltage vectors is calculated according to formula (1): d |V i and Δi q |V i , where V i Specifically, V1, V4, V5, V8, V9, V 12 、V 25 ; Among them, β di , β qi Represents the voltage vector V i The angle between the d-axis and the q-axis; β dk , β qk Represents the voltage vector V k The angle between the d-axis and the q-axis; δi d |V k ,δi q |V k Represents the voltage vector V k The forced response of the dq axis current under the action of The parameter values ​​are shown in Table 2; Table 2 In the table, V x Represents the amplitude coefficient between characteristic voltage vectors; If V k If it is a large vector, the basic voltage vector V in its voltage set is calculated according to formula (2): m The corresponding dq axis current difference Δi d |V m and Δi q |V m Then, according to formula (3), the dq axis current difference Δi corresponding to the basic voltage vector in the set of all voltage vectors is calculated: d |V i and Δi q |V i ; Among them, δi d |V m and δi q |V m They are voltage vector V m The forced response of dq axis current under the action; If V k Belongs to the middle vector and is the basic vector, then directly calculate the dq axis current difference Δi corresponding to the basic voltage vector in the collection of all voltage vectors according to formula (1): d |V i and Δi q |V i ; If V k It belongs to the middle vector but not to the basic vector. Then, the basic voltage vector V in the voltage set to which it belongs is calculated according to formula (4): n The corresponding dq axis current difference Δi d |V n and Δi q |V n Then, according to formula (5), the dq axis current difference Δi corresponding to the basic voltage vector in the set of all voltage vectors is calculated: d |V i and Δi q |V i ; Among them, δi d |V n and δi q |V n They are voltage vector V n The forced response of the dq axis current under the action of.

2. The improved three-level model-free predictive current control method for a permanent magnet synchronous motor according to claim 1, characterized in that: In step 2, the method for implementing the updating of the current difference lookup table is as follows: after obtaining the dq axis current difference corresponding to the basic voltage vector of the voltage set 1, 3, and 5, the dq axis current difference corresponding to other voltage vectors in the same set is calculated by formula (6); the dq axis current difference corresponding to all other voltage vectors of the voltage set 2, 4, and 6 is calculated by formula (7); the calculated dq axis current difference under different switching states is stored in a current difference lookup table containing 27 different switching states and the original data of the same switching state in the table is replaced to complete the updating of the current difference lookup table; Among them, parameter a takes 1, 3, 5, and parameter b takes 2, 4, 6, corresponding to the voltage vector number.

3. The improved three-level model-free predictive current control method for permanent magnet synchronous motor according to claim 1, characterized in that: The step 3 specifically includes: firstly obtaining the dq axis current difference Δi under different switching states from the current difference lookup table: d |V j and Δiq|V j ; Then, according to formula (8), the predicted current value i of the dq axis under different switching states at time (k+1) is calculated d (k+1)|V j and i q (k+1)|V j ; 4. The improved three-level model-free predictive current control method for a permanent magnet synchronous motor according to claim 1, characterized in that: In step 4, the vector candidate rule is as follows: consider the basic vector in the adjacent sector of the action vector at the previous sampling moment as the candidate vector, take the second principle that only one phase switch state changes continuously in one sampling cycle, screen small vectors, medium vectors, and large vectors, and then complete the small, medium, large, and zero vector types of the screened vectors, and finally the number of candidate vectors is 6-9; Table 3 Table 3 shows the final candidate rules.

Images