Permanent magnet motor model-free predictive current control method based on backward cuppman operator

Abstract

This invention discloses a model-free predictive current control method for permanent magnet synchronous motors based on the backward Koopman operator. The method involves processing the stator current vector of the permanent magnet synchronous motor, constructing an extended state vector equation, establishing a data-driven model for the conventional Koopman operator based on the extended state vector equation, and then establishing a data-driven model for the backward Koopman operator. The model is then adjusted and reorganized to obtain the coefficient matrix. Next, coefficient recursive vectors and data recursive vectors are constructed based on the coefficient matrix. The least squares method is used to estimate the coefficient recursive vectors and data recursive vectors to obtain the desired data. k The data recursive vector at time +1 is used to predict the stator voltage vector using time-shifting. The stator voltage vector is then transmitted to the modulation module, where it is converted into a switching signal and output to the inverter. This solves the problem of limited adaptability and inability to reflect complex motion characteristics in traditional model-free predictive control.

Description

Technical Field

[0001] This invention relates to the field of motor control, and more specifically to a model-free predictive current control method for permanent magnet motors based on the backward Koopman operator. Background Technology

[0002] Permanent magnet synchronous motor ( Permanent Magnet Synchronous Motor, PMSM Due to its high power density, high efficiency, and excellent speed regulation performance, it has been widely used in many fields such as new energy vehicles, high-end manufacturing, and aerospace. To achieve... PMSM High-performance control requires fast and precise closed-loop control of the stator current, the source of the core electromagnetic torque. Therefore, the control performance of the current loop directly determines the dynamic response, steady-state accuracy, and robustness of the entire drive system.

[0003] Among many advanced control strategies, model predictive control (MMC) is one of them. Model Predictive Control, MPC MPC has attracted much attention due to its intuitive concept and ease of handling multivariable constraint problems. However, traditional MPC relies on the mathematical model of the controlled object and the accuracy of its parameters. In practical applications, changes in the motor's operating environment and operating conditions will lead to changes in motor parameters. Coupled with inaccurate parameter measurements, these factors will affect the performance of the control algorithm and cause noise, static current errors, or even divergence during motor operation, thus affecting the control performance and stability of the system.

[0004] To overcome the inherent drawback of model dependence, model-free predictive control ( Model-Free Predictive Control, MFPC or model-less predictive current control ( Model-Free Predictive Current Control, MFPCC Model-free predictive control (MDC) has emerged as a new technology. The core concept of this approach is to abandon traditional mathematical models of motors and instead construct a data-driven predictive model directly from the system's input-output data. Typical MDC methods directly collect stator current values ​​(state variables) and control voltages from the current and several past moments, extrapolating future current values ​​by calculating the rate of change or acceleration of the current, and using this as a predictive model for rolling optimization. However, MDC methods cannot consider the influence of higher-order or other complex motion characteristics of state variables on the system response, resulting in models with limited adaptability and an inability to accurately reflect the operating state and characteristics of the controlled object (i.e., the motor system).

[0005] In view of this, this application has conducted in-depth research, which led to the creation of this case. Summary of the Invention

[0006] The application aims to provide a permanent magnet motor model-free predictive current control method based on a backward Kupman operator, which can consider the high-order characteristics of state variables in the modeling process, obtain a data model with better adaptive performance, maintain the inherent robustness of the system, and improve the prediction accuracy and control performance.

[0007] To achieve the above-mentioned purpose, the solution of the application is:

[0008] A permanent magnet motor model-free predictive current control method based on a backward Kupman operator, three-phase stator currents and rotor position angles of a permanent magnet synchronous motor are obtained to obtain a stator current vector, then an extended state vector formula is constructed, a data-driven model of a conventional Kupman operator is established based on the extended state vector formula, a data-driven model of a backward Kupman operator is established, then the data-driven model of the backward Kupman operator is adjusted and reorganized to obtain a coefficient matrix, then a coefficient recursive vector and a data recursive vector are constructed according to the coefficient matrix; the least square method is used to estimate the coefficient recursive vector and the data recursive vector, the data recursive vector at time t+1 is obtained, and a time-shifted predicted stator voltage vector is used, then the obtained stator voltage vector is transmitted to a modulation module in a driving system of the permanent magnet synchronous motor, the modulation module is converted into a switching signal output to an inverter on the driving system for control. k +1time, and a time-shifted predicted stator voltage vector is used, then the obtained stator voltage vector is transmitted to a modulation module in a driving system of the permanent magnet synchronous motor, the modulation module is converted into a switching signal output to an inverter on the driving system for control.

[0009] Comprising the following steps:

[0010] Step 1, obtaining a stator current vector, three-phase phase currents and rotor position angles of the motor are obtained by sampling, the three-phase phase currents and the rotor position angles are all subjected to coordinate transformation to form a stator current vector, the stator current vector is i s ( k ), i s ( k ) =[ i sd ( k ), i sq ( k )] T , i sd ( k ) and i sq ( k ) are respectively k time d and q axis stator current components;

[0011] Step 2, constructing an extended state vector formula, constructing kThe expansion state vector at time t;

[0012] Step 3: Establish the data-driven model of the Koopman operator as follows:

[0013] ,

[0014] In the formula, for k The expanding state vector at time t, for The expanding state vector at time t, p The current calculation is at the th p At that moment, p The range of values ​​is , For integer mathematical symbols, N The highest power of the extended state vector; A p The extended state matrix contains A 1 、...、A N ; For the first i The input matrix for each prediction period contains B 1,0 ... B N,N-1 ; C p The output matrix contains C 1、...、 C N ; for k + i The stator voltage vector at time t. for k + p The stator current vector at time t;

[0015] Step 4: Construct a data-driven model for the backward Koopman operator, and... p Setting it to 1, and simultaneously shifting the extended state vector, the stator current vector, and the stator voltage vector backward in time, the resulting data-driven model for the backward Koopman operator is as follows.

[0016] ,

[0017] In the formula, for k The expansion state vector at time +1, for k - i The stator voltage vector at time t. fork The stator current vector at time +1; where, A 1 represents the first extended state matrix. B i For the first moment i Input matrix for each prediction period, C 1 represents the first output matrix;

[0018] Step 5: Construct the coefficient matrix, adjust and reorganize the data-driven model of the backward Koopman operator, and obtain the following formula.

[0019] ,

[0020] In the formula, for k The stator current vector at time t. For the reorganization k+ The extended state matrix at time 1, For the reorganization k The expanded state matrix at time t, For the first time after the reorganization i The input matrix for each prediction period; wherein, the coefficient matrix is,

[0021] ,

[0022] ,

[0023] ,

[0024] In the formula, for k The expansion state coefficients at time +1 for k The expansion state coefficient at time t, For the first i Input coefficients for each prediction period; For the current moment, the first i Input matrix for each prediction period;

[0025] Step 6: Construct a coefficient recursive vector and a data recursive vector, combining the coefficients in the coefficient matrix to form... k The recursive vector of time coefficients is ,

[0026] The expansion state vector, the stator current vector, and the stator voltage vector are combined to form... k The recursive vector of time data is ,

[0027] In the formula, They are respectively kFrom the moment on, until k - N Stator voltage vector at time +1 For stator current N Power;

[0028] Step 7: Estimate using the least squares model. Substitute the coefficient recursive vector and the data recursive vector from Step 6 into the least squares model as follows, and estimate the data recursive vector. The constructed least squares model is as follows:

[0029] ,

[0030] In the formula, for k Voltage error value at any time for k The recursive gain matrix at each time step, for k Covariance matrix at time -1 for k Time-varying covariance matrix for k -1 time coefficient recursive vector Forgetting factor;

[0031] Then, move the data recursive vector to k +1 time, constitutes k The recursive vector of the data at time +1 is,

[0032] ,

[0033] In the formula, This is the reference value for the stator current. The Nth power of the stator current reference value;

[0034] Step 8: Use time-shift prediction to determine the stator voltage vector. The formula for this predicted stator voltage vector is as follows: In the formula, for k Recursive vector of time coefficients.

[0035] In step 1, a current sensor is used to acquire the three-phase phase current of the permanent magnet synchronous motor, and an encoder is used to acquire the rotor position angle of the permanent magnet synchronous motor. The rotor position angle is... θ .

[0036] The three-phase phase currents in the three-phase coordinate system are converted into stator currents in the two-phase stationary coordinate system. Then, using the rotor position angle, the stator currents in the two-phase stationary coordinate system are converted into stator currents through... Park Transform to a rotating coordinate system to obtaink time d and q The stator current components of the shaft constitute the stator current vector.

[0037] In step 2, the constructed k The state vector expression at time t is:

[0038] ,

[0039] In the formula, for k The electron current vector at any given moment, This is an extended family of state equations for the stator current. The extended state equation for the first stator current is... middle This is a substitution symbol.

[0040] In step 5, the data-driven model of the backward Koopman operator is adjusted and reorganized into an external input autoregressive form, yielding the following adjustment formula: Then, through the equivalent transformation of the adjustment formula, the following formula is obtained. .

[0041] The drive system includes the modulation module and the inverter.

[0042] With the above structure, the present invention has the following beneficial effects: by introducing the backward Koopman operator on the basis of model-free predictive current control, the modeling process considers the higher-order characteristics of state variables, and uses the extended state as a direct cause to establish a mathematical model to characterize higher-order motion characteristics, thereby obtaining a data-driven model with better adaptability. In this way, while maintaining the inherent robustness of the system, the prediction accuracy and control performance are improved, and the problems of limited adaptability and difficulty in reflecting complex motion characteristics of traditional model-free predictive control are solved. Attached Figure Description

[0043] Figure 1 This is a block diagram illustrating the control principle of model-free predictive current control in this invention.

[0044] Figure 2 This is a structural diagram of the data-driven model based on the conventional Koopman operator in this invention.

[0045] Figure 3 This is a structural diagram of the data-driven model based on the backward Koopman operator in this invention.

[0046] Figure 4 This is a flowchart of the control method in this invention. Detailed Implementation

[0047] To further explain the technical solution of the present invention, the present invention will be described in detail below through specific embodiments.

[0048] A model-free predictive current control method for permanent magnet synchronous motors based on the backward Koopman operator is presented. This method utilizes a common drive system for controlling permanent magnet synchronous motors. The drive system includes an inverter and a modulation module. The inverter can be a conventional inverter, such as a two-level three-phase inverter, and includes several switching transistors. The modulation module is a conventional modulation module used to convert voltage signals (such as stator voltage vectors) into switching signals. It should be noted that the following description of outputting switching signals to the inverter to enable or disable the individual switching transistors in the inverter is a conventional technique.

[0049] In this embodiment, the permanent magnet synchronous motor is a commercially available permanent magnet synchronous motor, and the inverter is a conventional inverter.

[0050] like Figures 1-4 As shown, the model-free predictive current control method for a permanent magnet synchronous motor is as follows: The three-phase stator current and rotor position angle of the permanent magnet synchronous motor are obtained. Based on these, the stator current vector is obtained, and an extended state vector equation is constructed. Then, a data-driven model of the conventional Koopman operator is established based on the extended state vector equation, thereby establishing a data-driven model of the backward Koopman operator. The backward Koopman operator data-driven model is then adjusted and reorganized to obtain the coefficient matrix. Next, coefficient recursion vectors and data recursion vectors are constructed based on the coefficient matrix. The least squares method is used to estimate the coefficient recursion vectors and data recursion vectors to obtain the desired data. k The data recursive vector at time +1 is used to predict the stator voltage vector using time shift. The obtained stator voltage vector is transmitted to the modulation module, which converts it into a switching signal and outputs it to the inverter to control the on / off state of each switching transistor in the inverter.

[0051] To elaborate, the specific steps include the following.

[0052] Step 1: Obtain the stator current vector i s ( k The three phases of the permanent magnet synchronous motor (i.e., ) are obtained through sampling. A , B , C (Three-phase) phase current and rotor position angle, where the rotor position angle is... θ The three-phase currents and rotor position angles are transformed using conventional coordinate methods to obtain the stator current vector, which is: i s ( k ),in, i s ( k) =[ i sd ( k ), i sq ( k )] T , i sd ( k )and i sq ( k ) are respectively k time d and q The stator current component of the shaft.

[0053] In this embodiment, a current sensor is used to detect the three-phase current of the permanent magnet synchronous motor, and an encoder is used to detect the rotor position angle of the permanent magnet synchronous motor. The sampling of the permanent magnet synchronous motor by the current sensor and the encoder, as well as their installation positions, are standard practices in existing motors and will not be described further.

[0054] Furthermore, based on the obtained three-phase currents and rotor position angles, the stator current vector is obtained through coordinate transformation, which is a conventional operation. The specific steps are as follows: the three-phase currents in the three-phase coordinate system are transformed... Clark The transformation matrix is ​​used to convert the stator current in a two-phase stationary coordinate system. Then, the rotor position angle is used to convert the stator current in the two-phase stationary coordinate system through... Park Transform to a rotating coordinate system to obtain k time d and q The stator current component of the shaft, i.e. i sd ( k )and i sq ( k ), and constitute the stator current vector. i s ( k ) .

[0055] Step 2: Construct the extended state vector.

[0056] Construct in a rotating coordinate system k The expansion state vector at time t is:

[0057] (1),

[0058] In the formula, for k The expanding state vector at time t, for k The electron current vector at any given moment, This is an extended family of state equations for the stator current. The extended state equation for the first stator current is... For the first N The expanded state equations for the stator currents, middle It is the transpose symbol. For stator current N The powers of 1; where the corresponding parameters in the ellipsis above are deduced in the same way as above, and will not be elaborated here.

[0059] Step 3: Establish the data-driven model of the conventional Koopman operator: Based on the extended state vector formula in Step 2, establish the data-driven model of the conventional Koopman operator as follows.

[0060] (2),

[0061] In the formula, for The expanding state vector at time t, p The current calculation is at the th p At that moment, p The range of values ​​is , For integer mathematical symbols, N The highest power of the extended state vector; A p The extended state matrix contains A 1 、...、A N ; For the first i The input matrix for each prediction period contains B 1,0 ... B N,N-1 ; C p The output matrix contains C 1、...、 C N ; Solving for the required prediction k + i The stator voltage vector at time t is the dependent variable that needs to be solved. for k + p The stator current vector at time t.

[0062] It should be noted that the above-mentioned conventional Koopman operator predicts backward from the current time and cannot directly characterize and train the temporal characteristics of the sampled data. Therefore, it cannot be directly applied to the next time observation function required in this embodiment. Therefore, it is necessary to proceed with step 4.

[0063] Step 4: Construct the data-driven model of the backward Koopman operator. For real-time control systems, an excessively long prediction time domain will lead to the accumulation of prediction errors, which is detrimental to control performance optimization. Therefore, the prediction time domain is limited, i.e. p Setting it to 1, and simultaneously shifting the extended state vector, stator current vector, and stator voltage vector backward in time, the data-driven model of the backward Koopman operator is obtained as follows:

[0064] (3),

[0065] In the formula, for k The expansion state vector at time +1, for k - i The stator voltage vector at time t. for k The stator current vector at time +1; where, since the p value is set, the above matrices are compressed, therefore in equation (3), A 1 represents the first extended state matrix. B i For the first moment i Input matrix for each prediction period, C 1 represents the first output matrix.

[0066] It should be noted that, as Figures 3-4 As shown, due to p Fixed at 1, and with a backward time shift, the accumulated stator voltage vector from k Time becomes k- ( N -1), therefore in formula (2) p -1 changes and transforms into N -1.

[0067] Step 5: Construct the coefficient matrix. To achieve a continuous control set, the architecture of the data-driven model of the backward Koopman operator is adjusted and reorganized to become an external input autoregressive model, resulting in the following formula: In the formula, for k The stator current vector at time t. For the reorganization k The expanded state matrix at time +1, For the reorganization k The expanded state matrix at time t, For the first time after the reorganization i Input matrix for each prediction period.

[0068] To elaborate, in formula (4), we can perform an equivalent transformation from formula (3), and the derivation process is as follows:

[0069] (4).

[0070] The coefficient matrix is ​​as follows:

[0071] (5),

[0072] (6),

[0073] (7),

[0074] In the formula, for k+ The expansion state coefficients at time 1 for k The expansion state coefficient at time t, For the first i Input coefficients for each prediction period; For the current time (i.e., the 0th time), the... i Input matrix for each prediction period, This is the first extended state matrix.

[0075] Step 6: Construct the coefficient recursive vector and the data recursive vector.

[0076] The coefficient matrix in step 5 , and The combination of coefficients in the formula constitutes k The time coefficient recursive vector, k The formula for the recursive vector of time coefficients is:

[0077] (8);

[0078] The expanded state vector, stator current vector, and stator voltage vector are combined to form... k The recursive vector of time data, k The recursive vector of time data is: (9);

[0079] In the formula, They are respectively k From the moment on, until k - N Stator voltage vector at time +1 For stator current N Power of 1.

[0080] Step 7: Use the least squares model for estimation. Substitute formulas (8) and (9) into the least squares model constructed below, and estimate the data recursion vector. The least squares model is as follows:

[0081] (10),

[0082] In the formula, for k Voltage error value at any time for k The recursive gain matrix at each time step, for k Covariance matrix at time -1 for k Time-varying covariance matrix for k -1 time coefficient recursive vector is the forgetting factor; where the coefficient recursive vector here needs to be estimated for model updates.

[0083] Furthermore, since the sampling period in a permanent magnet synchronous motor is typically in the microsecond range, shifting the data recursive vector to... k +1 time, which constitutes k The recursive vector of data at time +1, k The data recursive vector at time +1 is:

[0084] (11), where, This is the reference value for the stator current. The value is the Nth power of the stator current reference value, meaning that this reference value will be used for all future extended states. replace.

[0085] Step 8: Use time-shift prediction to determine the stator voltage vector. The formula for predicting the stator voltage vector is as follows: (12), where, for k Recursive vector of time coefficients.

[0086] Step 9: The stator voltage vector obtained in Step 8 is output to the modulation module, which converts it into a switching signal and outputs the switching signal to the frequency converter to control the on / off state of each switching transistor in the frequency converter.

[0087] This invention discloses a model-free predictive current control method for permanent magnet motors based on the backward Koopman operator. By introducing the backward Koopman operator, it fully exploits the high-order and complex motion characteristics inherent in the input-output data of the motor system, constructing a data-driven predictive model with strong adaptability. Specifically, the method first obtains the stator current vector and rotor position angle of the motor, constructs an expanded state vector after coordinate transformation, and establishes a conventional Koopman operator model based on this. Then, it obtains the data-driven model of the backward Koopman operator through backward time shift. Subsequently, the model parameters are combined into coefficient recursive vectors and data recursive vectors, and online estimation is performed using the least squares method to predict the current at future times. Based on this, the stator voltage vector is obtained through time shift prediction, and converted into a switching signal to control the inverter via a modulation module. Thus, without relying on a precise mathematical model of the motor, it effectively improves the dynamic response speed, control accuracy, and operational stability of the system, and is particularly suitable for complex operating conditions such as time-varying motor parameters and load disturbances. This method not only overcomes the limitations of traditional model-free predictive control in that it does not adequately consider the motion characteristics of state variables, but also enhances the ability to represent the complex dynamic behavior of motor systems through data-driven modeling, while maintaining the inherent robustness of the system.

[0088] The above description is only a preferred embodiment of this invention. Any equivalent changes and modifications made within the scope of the claims of this invention shall fall within the scope of the claims of this invention.

Claims

1. A permanent magnet motor model-free predictive current control method based on backward Kuppman operator, characterized in that: Three-phase stator currents and rotor position angles of a permanent magnet synchronous motor are acquired to obtain a stator current vector, then an extended state vector formula is constructed, a data-driven model of a conventional Kuppmann operator is established based on the extended state vector formula, a data-driven model of a backward Kuppmann operator is established, then the data-driven model of the backward Kuppmann operator is adjusted and reorganized to obtain a coefficient matrix, then a coefficient recursive vector and a data recursive vector are constructed according to the coefficient matrix; the coefficient recursive vector and the data recursive vector are estimated by using a least square method, data recursive vectors at time points of +1 and +2 are obtained, and a stator voltage vector is predicted by using time shift prediction, then the obtained stator voltage vector is transmitted to a modulation module in a driving system of the permanent magnet synchronous motor, the modulation module is converted into a switching signal output to an inverter on the driving system for control; k +1 and +2 are obtained, and a stator voltage vector is predicted by using time shift prediction, then the obtained stator voltage vector is transmitted to a modulation module in a driving system of the permanent magnet synchronous motor, the modulation module is converted into a switching signal output to an inverter on the driving system for control; The method comprises the following steps: step 1, obtaining three-phase stator currents and rotor position angles of a permanent magnet synchronous motor to obtain a stator current vector; Step 2, Constructing the extended state vector equation, constructing in the rotating coordinate system k the extended state vector equation at the time instant Step 3, a data-driven model of a conventional Kuppmann operator is established as follows, ， wherein is the expanded state vector at time instant k is the expanded state vector at time instant is the expanded state vector at time instant is the expanded state vector at time instant p is the current computation at the p th time instant, p is the value range of , is the integer mathematical symbol, N is the highest dimension of the expanded state vector; A p is the expanded state matrix, containing A 1 、...、A N ; is the i th prediction period input matrix, containing B 1,0 ,..., B N,N-1 ; C p is the output matrix, containing C 1,..., C N ; is the k + i stator voltage vector at time instant is the k + p stator current vector at time instant Step 4, constructing a data-driven model of the backward Kuppman operator, yielding p is set to 1, while the extended state vector, the stator current vector and the stator voltage vector are all backward time-shifted, the data-driven model of the backward Kuppman operator is as follows, ， In the formula, B i is the input matrix of the first prediction cycle at the first time instant; and i is the input matrix of the second prediction cycle at the first time instant. Step 5, a coefficient matrix is constructed, the data-driven model of the backward Kuppmann operator is adjusted and reorganized, and the following formula is obtained, ， wherein is k the stator current vector at the time instant is the reorganized k the expanded state matrix at the time instant is the reorganized k the expanded state matrix at the time instant is the reorganized i input matrix for the prediction period; wherein the coefficient matrix is ， ， ， wherein is k+ is the dilated state coefficient at time is k is the dilated state coefficient at time is the input coefficient of the i th prediction period; is the input matrix of the i th prediction period at the current time; Step 6, constructing the coefficient recursive vector and the data recursive vector, combining each coefficient in the coefficient matrix to form k the coefficient recursive vector at time t is combining the expanded state vector, the stator current vector and the stator voltage vector to form k the data recursive vector at time t is , In the formula, respectively k from the moment k - N the stator voltage vector at time +1, is the η power, where the highest power η is N , that is, consistent with the dimension; Step 7, estimation is performed by using a least square method model, the coefficient recursive vector and the data recursive vector in step 6 are substituted into the following least square method model, and estimation is performed on the data recursive vector, wherein the least square model constructed is as follows, ， wherein is k the voltage error value at time instant is k the recursive gain matrix at time instant is k the covariance matrix at time instant is k the covariance matrix at time instant is k the coefficient recursive vector at time instant is the forgetting factor; Then, the data recursive vector is moved to k At time +1, the data recursive vector is k At time +1, the data recursive vector is ， In the formula, is a stator current reference value, is a stator current reference value η power, wherein the highest power η is taken as N ; Step 8: Use time-shift prediction to determine the stator voltage vector. The formula for this predicted stator voltage vector is as follows: In the formula, for k Recursive vector of time coefficients.

2. The PM machine model-free predictive current control method based on backward Kuppman operator according to claim 1, characterized in that: In step 1, three-phase phase currents and a rotor position angle of the motor are acquired by sampling, the three-phase phase currents and the rotor position angle are all coordinate-transformed to constitute a stator current vector, the stator current vector is i s ( k ), i s k i sd k i sq k T , i sd k and i sq k are stator current components of the k moment d and q axis respectively.​​​​​​​​ 3. The PM machine model-free predictive current control method based on backward Kuppman operator according to claim 2, characterized in that: In step 1, a current sensor is used to obtain the three-phase phase currents of the permanent magnet synchronous motor, and an encoder is used to obtain the rotor position angle of the permanent magnet synchronous motor, wherein the rotor position angle is θ .

4. The PM machine model-free predictive current control method based on backward Kuppman operator according to claim 3, characterized in that: The three-phase phase currents in a three-phase coordinate system are converted into stator currents in a two-phase stationary coordinate system, and then the rotor position angle is adopted to convert the stator currents in the two-phase stationary coordinate system into a rotating coordinate system to obtain Park k the stator current components in the d-axis and the q-axis at the moment d and q and constitute the stator current vector.​ 5. The PM machine model-free predictive current control method based on backward Kuppman operator according to claim 1, characterized in that: In step 2, the constructed k The state vector equation at time instant t is ， wherein is k the instantaneous electronic current vector, is the extended state equation cluster of stator currents, is the first extended state equation of stator currents, in is the permutation symbol.

6. The PM machine model-free predictive current control method based on backward Kuppman operator according to claim 1, characterized in that: In step 5, the data-driven model of the backward Kuppers operator is adjusted and reorganized into an external input autoregressive form, and the following adjustment formula is obtained Then, the following formula is obtained through equivalent transformation of the adjustment formula .

7. The model-free predictive current control method for permanent magnet machines based on backward Kuppman operator according to any of claims 1-6, characterized in that: The drive system comprises the modulation module and the inverter.

Images