A self-adaptive current control parameter setting method for permanent magnet synchronous motor

Abstract

The application discloses a permanent magnet synchronous motor adaptive current control parameter setting method, comprising the following steps: S1, establishing a mathematical model of a five-phase permanent magnet synchronous motor; S2, designing a model-free adaptive controller and a pseudo-block Jacobi matrix estimation algorithm; and S3, setting the model-free adaptive current control parameters offline based on a central collision optimization algorithm. The application realizes stable, fast and robust control performance in a five-phase permanent magnet synchronous motor driving system.

Description

Technical Field

[0001] This invention relates to the field of model-free adaptive current control (MFACC) for five-phase permanent magnet synchronous motors, specifically to a method for tuning adaptive current control parameters for permanent magnet synchronous motors. Background Technology

[0002] In recent years, with the continuous development of green energy, permanent magnet synchronous motors (PMSMs) have attracted much attention in various application fields such as electric vehicles, ship electric propulsion, rail transportation, and low-altitude aircraft. Compared with traditional three-phase permanent magnet synchronous motors, five-phase permanent magnet synchronous motors have stronger fault tolerance, harmonic subspace suppression performance, and higher power density. To meet the stringent high-performance requirements of servo systems, researchers have conducted extensive research on various control strategies, including field-oriented control (FOC), direct torque control (DTC), and model predictive control (MPC).

[0003] In industrial applications, Field-Oriented Control (FOC) of permanent magnet synchronous motors is considered the mainstream control strategy and has been widely used over the past few decades. While proportional-integral (PI) controllers in field-oriented control are widely recognized for their simple structure and ease of implementation, they require extensive parameter tuning. Furthermore, modeling errors in the transfer function of the motor drive system often affect the accuracy of PI parameter tuning. Directed Turbocharging (DTC) offers advantages such as simple structure, rapid dynamic response, and low sensitivity to parameter changes. However, because this method relies on hysteresis comparators and switching meters to achieve torque and stator flux linkage control, the system suffers from steady-state torque errors and significant torque ripple. Multi-Purpose Control (MPC), relying on precise mathematical models, can effectively handle multiple constraints and control objectives, making it suitable for complex nonlinear systems. However, its control performance degrades significantly under parameter mismatch conditions. Deadbeat Predictive Current Control (DPCC) is based on discrete-time mathematical models to calculate the voltage required for the next sampling time. This method can eliminate current tracking errors within one control cycle and achieve low harmonic distortion under rated parameter conditions. However, DPCC has high requirements for the accuracy of motor parameters, poor robustness, and does not have the ability to resist disturbances and handle constraints, so error compensation is required.

[0004] Model-free control, proposed by Fliess and Join, has gained widespread attention in various fields, including robot control, modern power systems, vehicle suspension systems, and permanent magnet synchronous motor (PMSM) drive systems. Unlike model predictive current control (MPCC), model-free predictive current control (MFPCC) does not rely on the physical parameters of the motor. Instead, it achieves drive control of the PMSM by employing a hyperlocal model or a dynamically linearized mathematical framework, combined with input-output data. Therefore, this method is classified as a type of data-driven predictive control method. Currently, many researchers have conducted extensive research on the application of model-free predictive current control in permanent magnet synchronous motor drive systems. Zhang et al. proposed a model-free predictive current control method based on an extended state observer (ESO). This method constructs a hyperlocal model and designs a linear extended state observer with current error feedback. This scheme can achieve robust motor control without requiring motor parameters, but the selection of the observer bandwidth ω0 and parameter α still needs to be empirically tuned according to specific operating conditions. Ma et al. proposed an improved model-free predictive current control strategy. This strategy utilizes the relationship between two consecutive applied voltage vectors and the resulting current change rate. By deriving the current change rate of all feasible voltage vectors within a control cycle, and then using a lookup table method to obtain the stator current to evaluate the value function, the optimal voltage vector is selected for the next sampling time.

[0005] Model-free adaptive control (MFAC), initially proposed by Hou, has been successfully applied in various fields such as agent learning control, power systems, and path tracking for autonomous vehicles. Based on the principle of compression mapping, Hou et al. conducted a rigorous theoretical analysis of a model-free adaptive control scheme based on full-format dynamic linearization (FFDL). This research proved the bounded input-bounded output stability, internal stability, and dynamic monotonic convergence of the tracking error of this method, and it has been successfully applied in motor drive systems. Furthermore, Wang et al. proposed a model-free adaptive current control (MFACC) method that does not rely on motor parameters. This method only uses historical dq-axis current and voltage data to estimate the dq-axis voltage at the next sampling time and uses a non-dominated sorting genetic algorithm II (NSGA-II) for parameter optimization, meeting the high-performance control requirements of three-phase permanent magnet synchronous motors. Aghaei Hashjin applied model-free adaptive control to the phase current control of wound-rotor synchronous motors. Compared with traditional proportional-integral (PI) control, the model-free adaptive control method generates additional steady-state oscillations.

[0006] Intelligent optimization algorithms are a class of metaheuristic optimization methods that simulate biological behavior, evolutionary processes, or physical laws. These algorithms can effectively solve complex optimization problems with high dimensionality, nonlinearity, and multimodal characteristics. By implementing iterative information interaction between individuals, the algorithm can autonomously search for the global optimum within a given constraint space. Typical algorithms include Particle Swarm Optimization (PSO), Grey Wolf Optimization (GWO), and Genetic Algorithm (GA). Center Collision Optimization (CCO) uses a center collision strategy as a unified position update rule, simulating elastic and inelastic particle collision mechanisms, and guides population updates with dominant individuals as the center. Compared with traditional and recently proposed algorithms, CCO shows significant advantages in convergence speed, optimization accuracy, and global search capability. Summary of the Invention

[0007] To address the aforementioned issues, this invention proposes a model-free adaptive current control (MFACC) parameter tuning method for a five-phase permanent magnet synchronous motor based on the center collision optimization algorithm (CCO). To reduce dependence on motor parameters, minimize the reliance on manual experience in parameter adjustment, and find the globally optimal parameter values, this method first performs full-format dynamic linearization (FFDL) processing on the sampled voltage and current data of the five-phase permanent magnet synchronous motor in a synchronous rotating coordinate system, constructs a pseudo-block Jacobian matrix (PPJM) estimation function, and designs the MFACC controller. Then, it selects some parameters, sets parameter limits, and designs the optimization objective for the motor drive. Subsequently, it uses the center collision optimization algorithm to determine the optimal parameters for model-free adaptive current control through offline iteration, effectively solving the parameter adjustment problem. Experimental results show that the proposed CCOMFACC scheme achieves stable, fast, and robust control performance in a five-phase permanent magnet synchronous motor drive system.

[0008] The specific plan is as follows:

[0009] A method for tuning adaptive current control parameters of a permanent magnet synchronous motor includes the following steps:

[0010] S1. Establish a mathematical model for a five-phase permanent magnet synchronous motor.

[0011] Based on the topology of a five-phase permanent magnet synchronous motor driven by a two-level voltage source inverter, the phase bridge arm switching function is defined, 32 voltage vectors are generated and mapped to the α-β principal plane and the xy third harmonic plane. The third harmonic is eliminated by the nearest four vector space vector pulse width modulation (NFV-SVPWM) technology. The voltage transformation relationship between the stationary coordinate system and the synchronous rotating coordinate system is established, and the stator voltage equation, electromagnetic torque equation and mechanical motion equation of the d1-q1 axis and d3-q3 axis are obtained.

[0012] S2. Design a model-free adaptive controller and a pseudo-block Jacobian matrix estimation algorithm.

[0013] The five-phase permanent magnet synchronous motor system is represented as a multi-input multi-output nonlinear discrete-time system. Based on the full-format dynamic linearization FFDL, the nonlinear system is transformed into a linear data model. A model-free adaptive controller with cost function minimization is designed, and an estimation algorithm for the pseudo-block Jacobian matrix PPJM is constructed. Reset conditions are set to prevent parameter drift.

[0014] S3. Offline tuning of model-free adaptive current control parameters based on center collision optimization algorithm

[0015] The input and output variables of the model-free adaptive current controller are defined, the parameters to be optimized are determined, and an optimization function is established with the goal of suppressing the third harmonic current and reducing torque ripple. The center collision optimization algorithm is used to adaptively allocate the search ratio in the original space and the decorrelation space, and the control parameters are iteratively optimized. The optimal parameters are applied to the controller to realize the current control of the five-phase permanent magnet synchronous motor.

[0016] Further, step S1 includes:

[0017] S11, Mathematical Model of Five-Phase Permanent Magnet Synchronous Motor

[0018] In the topology of a five-phase permanent magnet synchronous motor (PMSM) system driven by a two-level voltage source inverter (2L-VSI), the phase voltage of each phase arm is directly determined by the switching state of the corresponding arm. The switching function of the p-th phase (p∈{a,b,c,d,e}) is defined as S. p ∈{0,1}: When the switch in the phase bridge arm is turned on, S p =1, when off, S p =0; Since there are two switching states for each phase arm, the 2L-VSI of the five-phase PMSM can generate 32 different combinations of switching signals; the mathematical expression of the phase voltage generated by these switching states is:

[0019] (1)

[0020] Among them, V p It is the stator phase voltage of phase p, V dc It is the DC bus voltage, S p This is the switching state of phase p;

[0021] The 2L-VSI based on the five-phase PMSM generates 32 switching states, corresponding to 32 voltage vectors. These vectors are mapped to the α-β principal plane and the xy third harmonic plane, forming vector distributions of different amplitudes. Based on the amplitude, these vectors in the α-β plane are divided into 10 large vectors, 10 medium vectors, and 10 small vectors, respectively marked by blue V... l The red-marked V mThe green-marked V s The display shows the corresponding voltage amplitudes as 0.6472Vdc, 0.4Vdc, and 0.2472Vdc, and two zero vectors; the voltage vectors in the α-β plane and the xy plane are denoted as V... αβ and V xy This voltage is obtained from the five-phase voltage transformation in the natural coordinate system and is expressed as:

[0022] (2)

[0023] Where j is the imaginary unit, and γ1 and γ3 are complex exponential rotation operators, corresponding to the phase offset coefficients of the fundamental and third harmonic components of the five-phase PMSM system, respectively.

[0024] To eliminate the third harmonic in a five-phase permanent magnet synchronous motor, the nearest four-vector space vector pulse width modulation (NFV-SVPWM) technology is adopted. Two large vectors and two medium vectors are selected, and the third harmonic is eliminated by allocating appropriate execution times. The α-β subspace is divided into 10 sectors. In the first sector, the selected vectors are V25, V24, V16, V29, and zero vectors V0 and V31. Each voltage vector is synthesized into a voltage vector of 0 in the xy subspace according to the execution times T1, T2, T3, and T4.

[0025] To achieve decoupled control of the stator voltage in the synchronous rotating coordinate system, the fundamental voltage vector V in the stationary coordinate system is... αβ and the third harmonic voltage vector V xy Transforming to the fundamental d1-q1 synchronous rotating coordinate system and the third harmonic d3-q3 synchronous rotating coordinate system respectively, the corresponding Park transformation relationships are expressed as follows:

[0026] (3)

[0027] Among them, V d1 V q1 V represents the d1-q1 axis voltage component in the fundamental subspace synchronous rotating coordinate system. d3 V q3 For the d3-q3 axis voltage components in the synchronous rotating coordinate system of the third harmonic subspace, θ e It is the electrical angle of the PMSM rotor;

[0028] The stator voltage equations for the five-phase permanent magnet synchronous motor along the d1-q1 and d3-q3 axes are expressed as follows:

[0029] (4)

[0030] Among them, R s It is the stator resistance, i d1i q1 i d3 i q3 These are the fundamental plane d1-q1 axis currents and the harmonic plane d3-q3 axis currents, respectively. d1 ,L q1 It is the stator inductance along the d1-q1 axis, L l It is the stator leakage inductance, D(·) is the time derivative, ω e It is the electric angular velocity of the rotor, ψ f It is the magnetic flux linkage of a permanent magnet; the electromagnetic torque of a five-phase PMSM is expressed as:

[0031] (5)

[0032] Among them, T e It is electromagnetic torque, P n This refers to the number of pole pairs of the motor; the mechanical motion equation of a five-phase PMSM is expressed as:

[0033] (6)

[0034] Where J is the moment of inertia, ω m It is the mechanical angular velocity of the rotor, T l It is the load torque, B m It is the coefficient of friction.

[0035] Further, step S2 includes:

[0036] This step introduces the full-format dynamic linearization (FFDL) of nonlinear systems, the design of the controller, and the pseudo-block Jacobian matrix (PPJM) for controller updates.

[0037] S21, FFDL of Multiple Input Multiple Output System

[0038] The multi-input multi-output nonlinear discrete-time system is represented as:

[0039] (7)

[0040] Where u(k) and y(k) are the input and output data at time k in the multiple-input multiple-output system, and kn u It is the nth u Historical input data, kn y It is the nth y Historical output data for each;

[0041] When designing a model-free adaptive current controller, a multi-input multi-output system must satisfy two assumptions:

[0042] Assumption 1: Except for finite time points, nonlinear functions , i=1 to i=ny The partial derivatives of the +2 variables all exist;

[0043] Assumption 2: Except at finite time points, the multiple-input multiple-output system satisfies the Lipschitz condition, that is, for any k1≠k2, k1,k2≥0 and u(k1)≠u(k2), it can be expressed as:

[0044] (8)

[0045] Where b is a constant greater than 0, L y and L u It is the length of the historical input / output data window, and satisfies 1≤L y ≤n y , 1≤L u ≤n u H Ly,Lu (k) is a length of L y +L u The information vector is represented as:

[0046] (9)

[0047] The system satisfies the above two assumptions, for any fixed L y and L u , ||ΔH Ly,Lu If (k)||≠0, then there exists a time-varying pseudo-block Jacobian matrix PPJM as Φ f,Ly,Lu (k) transforms the nonlinear system into a fully dynamically linearized FFDL data model, represented as:

[0048] (10)

[0049] Where, ΔH Ly,Lu (k)=H Ly,Lu (k)-H Ly,Lu (k-1) is the difference between the elements in the information vector, Φ f,Ly,Lu (k)=[φ1(k),…,φ Ly (k),φ Ly +1(k), …,φ Ly+Lu [(k)] is a PPJM matrix, and ||Φ f,Ly,Lu (k)|| is a bounded matrix, represented as:

[0050] (11)

[0051] Where b1 ≠ 0, b2 > 0, a > 1, m is the dimension of the input and output, satisfying b2 > b1(2a+1)(m-1), φ ij(Ly+1) It is an element in the (Ly+1)th square block;

[0052] S22, Model-Free Adaptive Controller Design

[0053] Based on the full-format dynamic linearized FFDL data model (10), the cost function of the controller is expressed as:

[0054] (12)

[0055] Among them, y * (k+1) is the desired output of the controller, and λ is a weighting factor. The nonlinear discrete-time system (7) is introduced into the cost function (12) so that the derivative of the cost function (12) with respect to u(k) is equal to 0, thus obtaining the minimum value of the cost function (12). Therefore, the controller is expressed as:

[0056] (13)

[0057] Where Δy(k) = y(k) - y(k-1) and Δu(k) = u(k) - u(k-1) are the differences between the input and output, and ρ i i=1,2,…,L y +L u It is L y +L u A pseudo-order weighting factor;

[0058] S23, PPJM estimation algorithm

[0059] Based on the full-format dynamic linearized FFDL data model (10), the cost function for pseudo-block Jacobian matrix estimation is expressed as:

[0060] (14)

[0061] in, yes The estimated value, μ is a weighting factor that makes the cost function (14) effective for... The derivative of is equal to 0, which gives the minimum value of the cost function (14). The pseudo-block Jacobian matrix estimate is expressed as:

[0062] (15)

[0063] To prevent parameter drift and meaningless estimation, it is necessary to... Set the reset conditions as follows:

[0064] (16)

[0065] (17)

[0066] Where a > 1, m is the system input / output dimension, η is the step size factor, μ is the weight factor, and b1 and b2 are two integers. yes The initial value.

[0067] Further, step S3 includes:

[0068] S31. Parameter optimization of model-free adaptive current control

[0069] When applying model-free adaptive current control (MFACC) to the current control of a five-phase PMSM mathematical model, preliminary parameter tuning of MFACC is required. First, the inputs and outputs y(k) and u(k) of the MFACC method are defined as follows:

[0070] (18)

[0071] Among them, i d1 (k), i q1 (k) and u d1 (k), u q1 (k) represents the historical current and voltage input data of the MFACC controller, defined as u d1 (k+1), u q1 (k+1) is the voltage output result of the MFACC controller;

[0072] Therefore, pre-fixing some parameters simplifies controller design. The input and output dimensions of the drive system are m=2. Weight parameters η and μ are selected in the pseudo-block Jacobian matrix, with a lower limit of b2=1e-6 and an upper limit of a*b2=50. Matrix elements are initialized. φ 11 , φ 22 , φ 12 and φ 21 Not setting b1 can reduce the number of elements φ. 12 and φ 21 Impact; Selecting the length L of historical data y =3 and L u =3; Select parameter ρ in controller u(k). i and λ, ρ i We use a single parameter ρ to replace the others; therefore, the center collision optimization algorithm needs to optimize 6 parameters, namely... ;

[0073] Select a set of optimization parameters =[0.3,80,-8,3,-0.5,-5] was used as the baseline. To study the influence of different parameters on the driving performance of the five-phase PMSM, these parameters were adjusted up and down relative to their rated values; since η and μ have a relatively small impact on driving performance, the analysis focused on [ρ,λ, φ 11 , φ 22 The effects of ];

[0074] The Center Collision Optimization (CCO) algorithm is inspired by the head-on collision equation in classical physics. This algorithm searches in both the original search space (OS) and the decorrelation space (DS). By adaptively allocating the search ratio between the two spaces, the algorithm achieves a balance between global exploration and local exploitation, effectively escaping local optima and accelerating convergence. The initial space allocation ratio is set at 50%:50%, and is dynamically adjusted based on the success rate of generating better solutions in the original and decorrelation spaces. In evaluations using the CEC 2017, CEC 2019, and CEC 2022 benchmark functions, the CCO algorithm demonstrates superior global search capability and local optimum avoidance ability. Based on limited data validation, the CCO algorithm can solve for the optimal parameters of the MFACC method, thereby improving the existing control performance of five-phase permanent magnet synchronous motors.

[0075] S32. Optimize the objective function

[0076] Suppressing harmonic currents and reducing torque ripple are important performance indicators for five-phase PMSM current control. When using the model reference adaptive control method under various operating conditions, uppercase J is selected as the optimization objective to quantify the control performance of the five-phase PMSM.

[0077] (19)

[0078] In the formula, N is the total number of sampling points in the motor control process, and λ J is the weighting coefficient of the third harmonic current, k is the current sampling point, and k∈[1,N]; the optimization objective J is constructed by comparing the actual d1-q1 axis current sampling value at the kth sampling point with its reference value, aiming to achieve real-time tracking of the actual d1-q1 axis current to the reference value and reduce the amplitude of the actual d3-q3 axis current;

[0079] S33. Offline MFACC parameter tuning method

[0080] The specific implementation steps of the MFACC method based on the CCO algorithm are as follows:

[0081] Step 1: Collect historical dq-axis voltage and current data of the five-phase PMSM in a synchronous rotating coordinate system, construct a data-driven model based on FFDL, design the MFACC controller, and establish an estimation algorithm for the PPJM; determine the lengths Ly and Lu of the historical data, and select the parameters of the MFACC controller. ;

[0082] Step 2: Set the experimental operating conditions for the five-phase PMSM; integrate the MFACC controller into the drive system of the five-phase PMSM, and set the parameters to be optimized. Integrate it into the CCO algorithm so that it can be called offline;

[0083] Step 3: Configure the initialization process of the CCO algorithm; determine the parameters. Given the constraints of the boundary conditions, design the objective optimization function J;

[0084] Step 4: Employ the CCO optimization algorithm to iteratively optimize the set of control parameters in step S3 offline, minimizing the objective function J, and outputting the optimal parameters. ;

[0085] Step 5: Apply the optimized parameters to the MFACC controller to achieve current control of the five-phase PMSM.

[0086] The beneficial effects of this invention are:

[0087] 1) The mathematical model of a five-phase permanent magnet synchronous motor is analyzed, and its dynamic linearization model is constructed based on dq voltage and current data. This method effectively solves the motor parameter mismatch problem using limited data.

[0088] 2) Set a reasonable optimization objective function and use the CCO algorithm to determine the optimal control parameters through offline iteration, thereby reducing the burden of manual tuning and ensuring the stability and optimality of the system.

[0089] 3) The robustness and feasibility of the proposed method are verified through comparative experiments under parameter mismatch, variable speed and variable torque conditions. Attached Figure Description

[0090] Figure 1 This is a diagram of the five-phase PMSM drive topology and winding configuration.

[0091] Figure 2 This is the basic vector distribution diagram of the 2L-VSI of the five-phase PMSM.

[0092] Figure 3 This is a composite diagram of NFV-SVPWM.

[0093] Figure 4 These are the effect diagrams of the optimization parameters. (a) ρ, (b) λ, (c) φ 11 ,(d) φ 22 .

[0094] Figure 5 This is the flowchart of the CCOMFACC control method.

[0095] Figure 6 These are experimental results of LESO under parameter mismatch conditions. (a) Phase current, (b) Current i in the αβ subplane. αβ (c) Current i in the xy subplane xy .

[0096] Figure 7 These are experimental results of the DPCC under parameter mismatch conditions. (a) Phase current, (b) Current i in the αβ subplane. αβ (c) Current i in the xy subplane xy .

[0097] Figure 8 These are experimental results of MFACC under parameter mismatch conditions. (a) Phase current, (b) Current i in the αβ subplane. αβ (c) Current i in the xy subplane xy .

[0098] Figure 9 These are experimental results of LESO under variable speed conditions. (a) Phase current, (b) Current i in the αβ subplane. αβ (c) Current i in the xy subplane xy .

[0099] Figure 10 These are experimental results of the DPCC under variable speed conditions. (a) Phase current, (b) Current i in the αβ subplane. αβ (c) Current i in the xy subplane xy .

[0100] Figure 11 These are experimental results of MFACC under variable speed conditions. (a) Phase current, (b) Current i in the αβ subplane. αβ (c) Current i in the xy subplane xy .

[0101] Figure 12 These are experimental results of LESO under variable torque conditions. (a) Phase current, (b) Current i in the αβ subplane. αβ (c) Current i in the xy subplane xy .

[0102] Figure 13These are experimental results of the DPCC under variable torque conditions. (a) Phase current, (b) Current i in the αβ subplane. αβ (c) Current i in the xy subplane xy .

[0103] Figure 14 These are experimental results of MFACC under variable torque conditions. (a) Phase current, (b) Current i in the αβ subplane. αβ (c) Current i in the xy subplane xy . Detailed Implementation

[0104] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0105] As shown in the figure, this invention provides a method for tuning adaptive current control parameters of a permanent magnet synchronous motor, comprising the following steps:

[0106] S1. Establish a mathematical model for a five-phase permanent magnet synchronous motor.

[0107] Based on the topology of a five-phase permanent magnet synchronous motor driven by a two-level voltage source inverter, the phase bridge arm switching function is defined, 32 voltage vectors are generated and mapped to the α-β principal plane and the xy third harmonic plane. The third harmonic is eliminated by the nearest four vector space vector pulse width modulation (NFV-SVPWM) technology. The voltage transformation relationship between the stationary coordinate system and the synchronous rotating coordinate system is established, and the stator voltage equation, electromagnetic torque equation and mechanical motion equation of the d1-q1 axis and d3-q3 axis are obtained.

[0108] S2. Design a model-free adaptive controller and a pseudo-block Jacobian matrix estimation algorithm.

[0109] The five-phase permanent magnet synchronous motor system is represented as a multi-input multi-output nonlinear discrete-time system. Based on the full-format dynamic linearization FFDL, the nonlinear system is transformed into a linear data model. A model-free adaptive controller with cost function minimization is designed, and an estimation algorithm for the pseudo-block Jacobian matrix PPJM is constructed. Reset conditions are set to prevent parameter drift.

[0110] S3. Offline tuning of model-free adaptive current control parameters based on center collision optimization algorithm

[0111] The input and output variables of the model-free adaptive current controller are defined, the parameters to be optimized are determined, and an optimization function is established with the goal of suppressing the third harmonic current and reducing torque ripple. The center collision optimization algorithm is used to adaptively allocate the search ratio in the original space and the decorrelation space, and the control parameters are iteratively optimized. The optimal parameters are applied to the controller to realize the current control of the five-phase permanent magnet synchronous motor.

[0112] In this embodiment, step S1 includes:

[0113] S11, Mathematical Model of Five-Phase Permanent Magnet Synchronous Motor

[0114] like Figure 1 As shown, in the topology of a five-phase permanent magnet synchronous motor (PMSM) system driven by a two-level voltage source inverter 2L-VSI, the phase voltage of each phase arm is directly determined by the switching state of the corresponding arm. The switching function of the p-th phase (p∈{a,b,c,d,e}) is defined as S. p ∈{0,1}: When the switch in the phase bridge arm is turned on, S p =1, when off, S p =0; Since there are two switching states for each phase arm, the 2L-VSI of the five-phase PMSM can generate 32 different combinations of switching signals; the mathematical expression of the phase voltage generated by these switching states is:

[0115] (1)

[0116] Among them, V p It is the stator phase voltage of phase p, V dc It is the DC bus voltage, S p This is the switching state of phase p;

[0117] The 2L-VSI based on the five-phase PMSM generates 32 switching states, corresponding to 32 voltage vectors; these vectors are mapped to the α-β principal plane and the xy third harmonic plane, forming vector distributions of different amplitudes, such as... Figure 2 As shown; based on their magnitudes, these vectors in the α-β plane are divided into 10 large vectors, 10 medium vectors, and 10 small vectors, respectively marked by the blue V... l The red-marked V m The green-marked V s The display shows the corresponding voltage amplitudes as 0.6472Vdc, 0.4Vdc, and 0.2472Vdc, and two zero vectors; the voltage vectors in the α-β plane and the xy plane are denoted as V... αβ and V xy This voltage is obtained from the five-phase voltage transformation in the natural coordinate system and is expressed as:

[0118] (2)

[0119] Where j is the imaginary unit, and γ1 and γ3 are complex exponential rotation operators, corresponding to the phase offset coefficients of the fundamental and third harmonic components of the five-phase PMSM system, respectively.

[0120] To eliminate the third harmonic in a five-phase permanent magnet synchronous motor, the nearest four-vector space vector pulse width modulation (NFV-SVPWM) technology is employed. Two large vectors and two medium vectors are selected, and by allocating appropriate execution times, the third harmonic is eliminated. The α-β subspace is divided into 10 sectors. In the first sector, the selected vectors are V25, V24, V16, V29, and the zero vectors V0 and V31. Each voltage vector, based on execution times T1, T2, T3, and T4, synthesizes a voltage vector of 0 in the xy subspace. Figure 3 As shown;

[0121] To achieve decoupled control of the stator voltage in the synchronous rotating coordinate system, the fundamental voltage vector V in the stationary coordinate system is... αβ and the third harmonic voltage vector V xy Transforming to the fundamental d1-q1 synchronous rotating coordinate system and the third harmonic d3-q3 synchronous rotating coordinate system respectively, the corresponding Park transformation relationships are expressed as follows:

[0122] (3)

[0123] Among them, V d1 V q1 V represents the d1-q1 axis voltage component in the fundamental subspace synchronous rotating coordinate system. d3 V q3 For the d3-q3 axis voltage components in the synchronous rotating coordinate system of the third harmonic subspace, θ e It is the electrical angle of the PMSM rotor;

[0124] The stator voltage equations for the five-phase permanent magnet synchronous motor along the d1-q1 and d3-q3 axes are expressed as follows:

[0125] (4)

[0126] Among them, R s It is the stator resistance, i d1 i q1 i d3 i q3 These are the fundamental plane d1-q1 axis currents and the harmonic plane d3-q3 axis currents, respectively. d1 ,L q1 It is the stator inductance along the d1-q1 axis, L l It is the stator leakage inductance, D(·) is the time derivative, ω e It is the electric angular velocity of the rotor, ψ f It is the magnetic flux linkage of a permanent magnet; the electromagnetic torque of a five-phase PMSM is expressed as:

[0127] (5)

[0128] Among them, T e It is electromagnetic torque, P n This refers to the number of pole pairs of the motor; the mechanical motion equation of a five-phase PMSM is expressed as:

[0129] (6)

[0130] Where J is the moment of inertia, ω m It is the mechanical angular velocity of the rotor, T l It is the load torque, B m It is the coefficient of friction.

[0131] In this embodiment, step S2 includes:

[0132] This step introduces the full-format dynamic linearization (FFDL) of nonlinear systems, the design of the controller, and the pseudo-block Jacobian matrix (PPJM) for controller updates.

[0133] S21, FFDL of Multiple Input Multiple Output System

[0134] The multi-input multi-output nonlinear discrete-time system is represented as:

[0135] (7)

[0136] Where u(k) and y(k) are the input and output data at time k in the multiple-input multiple-output system, and kn u It is the nth u Historical input data, kn y It is the nth y Historical output data for each;

[0137] When designing a model-free adaptive current controller, a multi-input multi-output system must satisfy two assumptions:

[0138] Assumption 1: Except for finite time points, nonlinear functions , i=1 to i=n y The partial derivatives of the +2 variables all exist;

[0139] Assumption 2: Except at finite time points, the multiple-input multiple-output system satisfies the Lipschitz condition, that is, for any k1≠k2, k1,k2≥0 and u(k1)≠u(k2), it can be expressed as:

[0140] (8)

[0141] Where b is a constant greater than 0, L y and L u It is the length of the historical input / output data window, and satisfies 1≤L y≤n y , 1≤L u ≤n u H Ly,Lu (k) is a length of L y +L u The information vector is represented as:

[0142] (9)

[0143] The system satisfies the above two assumptions, for any fixed L y and L u , ||ΔH Ly,Lu If (k)||≠0, then there exists a time-varying pseudo-block Jacobian matrix PPJM as Φ f,Ly,Lu (k) transforms the nonlinear system into a fully dynamically linearized FFDL data model, represented as:

[0144] (10)

[0145] Where, ΔH Ly,Lu (k)=H Ly,Lu (k)-H Ly,Lu (k-1) is the difference between the elements in the information vector, Φ f,Ly,Lu (k)=[φ1(k), …,φ Ly (k),φ Ly +1(k), …,φ Ly+Lu [(k)] is a PPJM matrix, and ||Φ f,Ly,Lu (k)|| is a bounded matrix, represented as:

[0146] (11)

[0147] Where b1 ≠ 0, b2 > 0, a > 1, m is the dimension of the input and output, satisfying b2 > b1(2a+1)(m-1), φ ij(Ly+1) It is an element in the (Ly+1)th square block;

[0148] S22, Model-Free Adaptive Controller Design

[0149] Based on the full-format dynamic linearized FFDL data model (10), the cost function of the controller is expressed as:

[0150] (12)

[0151] Among them, y *(k+1) is the desired output of the controller, and λ is a weighting factor. The nonlinear discrete-time system (7) is introduced into the cost function (12) so that the derivative of the cost function (12) with respect to u(k) is equal to 0, thus obtaining the minimum value of the cost function (12). Therefore, the controller is expressed as:

[0152] (13)

[0153] Where Δy(k) = y(k) - y(k-1) and Δu(k) = u(k) - u(k-1) are the differences between the input and output, and ρ i i=1,2,…,L y +L u It is L y +L u A pseudo-order weighting factor;

[0154] S23, PPJM estimation algorithm

[0155] Based on the full-format dynamic linearized FFDL data model (10), the cost function for pseudo-block Jacobian matrix estimation is expressed as:

[0156] (14)

[0157] in, yes The estimated value, μ is a weighting factor that makes the cost function (14) effective for... The derivative of is equal to 0, which gives the minimum value of the cost function (14). The pseudo-block Jacobian matrix estimate is expressed as:

[0158] (15)

[0159] To prevent parameter drift and meaningless estimation, it is necessary to... Set the reset conditions as follows:

[0160] (16)

[0161] (17)

[0162] Where a > 1, m is the system input / output dimension, η is the step size factor, μ is the weight factor, and b1 and b2 are two integers. yes The initial value.

[0163] In this embodiment, step S3 includes:

[0164] S31. Parameter optimization of model-free adaptive current control

[0165] When applying model-free adaptive current control (MFACC) to the current control of a five-phase PMSM mathematical model, preliminary parameter tuning of MFACC is required. First, the inputs and outputs y(k) and u(k) of the MFACC method are defined as follows:

[0166] (18)

[0167] Among them, i d1 (k), i q1 (k) and u d1 (k), u q1 (k) represents the historical current and voltage input data of the MFACC controller, defined as u d1 (k+1), u q1 (k+1) is the voltage output result of the MFACC controller;

[0168] Therefore, pre-fixing some parameters simplifies controller design. The input and output dimensions of the drive system are m=2. Weight parameters η and μ are selected in the pseudo-block Jacobian matrix, with a lower limit of b2=1e-6 and an upper limit of a*b2=50. Matrix elements are initialized. φ 11 , φ 22 , φ 12 and φ 21 Not setting b1 can reduce the number of elements φ. 12 and φ 21 Impact; Selecting the length L of historical data y =3 and L u =3; Select parameter ρ in controller u(k). i and λ, ρ i We use a single parameter ρ to replace the others; therefore, the center collision optimization algorithm needs to optimize 6 parameters, namely... ;

[0169] Select a set of optimization parameters =[0.3,80,-8,3,-0.5,-5] was used as the baseline. To study the influence of different parameters on the driving performance of the five-phase PMSM, these parameters were adjusted up and down relative to their rated values; since η and μ have a relatively small impact on driving performance, the analysis focused on [ρ,λ, φ 11 , φ 22 The effects and results are as follows: Figure 4 As shown:

[0170] The Center Collision Optimization (CCO) algorithm is inspired by the head-on collision equation in classical physics. This algorithm searches in both the original search space (OS) and the decorrelation space (DS). By adaptively allocating the search ratio between the two spaces, the algorithm achieves a balance between global exploration and local exploitation, effectively escaping local optima and accelerating convergence. The initial space allocation ratio is set at 50%:50%, and is dynamically adjusted based on the success rate of generating better solutions in the original and decorrelation spaces. In evaluations using the CEC 2017, CEC 2019, and CEC 2022 benchmark functions, the CCO algorithm demonstrates superior global search capability and local optimum avoidance ability. Based on limited data validation, the CCO algorithm can solve for the optimal parameters of the MFACC method, thereby improving the existing control performance of five-phase permanent magnet synchronous motors.

[0171] S32. Optimize the objective function

[0172] Suppressing harmonic currents and reducing torque ripple are important performance indicators for five-phase PMSM current control. When using the model reference adaptive control method under various operating conditions, uppercase J is selected as the optimization objective to quantify the control performance of the five-phase PMSM.

[0173] (19)

[0174] In the formula, N is the total number of sampling points in the motor control process, and λ J is the weighting coefficient of the third harmonic current, k is the current sampling point, and k∈[1,N]; the optimization objective J is constructed by comparing the actual d1-q1 axis current sampling value at the kth sampling point with its reference value, aiming to achieve real-time tracking of the actual d1-q1 axis current to the reference value and reduce the amplitude of the actual d3-q3 axis current;

[0175] S33. Offline MFACC parameter tuning method

[0176] The offline optimization flowchart of the MFACC method based on the CCO algorithm is as follows: Figure 5 As shown. The specific implementation steps of this offline optimization method are as follows:

[0177] Step 1: Collect historical dq-axis voltage and current data of the five-phase PMSM in a synchronous rotating coordinate system, construct a data-driven model based on FFDL, design the MFACC controller, and establish an estimation algorithm for the PPJM; determine the lengths Ly and Lu of the historical data, and select the parameters of the MFACC controller. ;

[0178] Step 2: Set the experimental operating conditions for the five-phase PMSM; integrate the MFACC controller into the drive system of the five-phase PMSM, and set the parameters to be optimized. Integrate it into the CCO algorithm so that it can be called offline;

[0179] Step 3: Configure the initialization process of the CCO algorithm; determine the parameters. Given the constraints of the boundary conditions, design the objective optimization function J;

[0180] Step 4: Employ the CCO optimization algorithm to iteratively optimize the set of control parameters in step S3 offline, minimizing the objective function J, and outputting the optimal parameters. ;

[0181] Step 5: Apply the optimized parameters to the MFACC controller to achieve current control of the five-phase PMSM.

[0182] Experimental results

[0183] The proposed CCOMFACC method was evaluated and its control performance in a five-phase surface-mounted permanent magnet synchronous motor drive system was verified through experiments. The parameter initialization and range of the CCO algorithm are shown in Table I. The hardware system of the experimental platform consisted of a real-time controller, oscilloscope, inverter, load motor, the five-phase PMSM under test, and a motor test bench. In the experiment, the system control frequency was set to 10 kHz, and the inverter dead time was set to 1 microsecond. To verify the control performance of the proposed method, the Linear Extended State Observer (LESO) method, Deadbeat Predictive Current Control (DPCC) method, and MFACC method were selected as comparative schemes. The comparative tests included motor parameter mismatch conditions, variable speed conditions, and variable torque conditions for the five-phase PMSM.

[0184] Table I: Parameters of the CCO Algorithm

[0185]

[0186] 1. Motor parameter mismatch condition

[0187] The aim is to verify the robustness of the motor algorithm under parameter mismatch conditions. During motor operation, the resistor R... s dq axis inductance L d L q and leakage inductance L l All parameters were set to 150% of the actual parameters to test the algorithm's performance. The motor operating condition was a step speed increase from 800 rpm to 2000 rpm, with 60% of the rated load torque applied. The results are as follows: Figure 6-9 As shown. From Figure 6As can be seen from (a), the THD of the A-phase current under the LESO method at 40%, 60%, 80%, and 100% rated speed are 50.27%, 48.69%, 47.11%, and 38.20%, respectively, and the five-phase current fluctuations are too large. Therefore, the LESO method requires high accuracy of its parameters. Figure 6 As can be seen from (b), i corresponds to the rotational speed of LESO. α The THDs of the currents were 50.3%, 48.64%, 47.08%, and 38.15%, respectively. Figure 6 As can be seen from (c), the harmonic current fluctuates by about 0.65A.

[0188] from Figure 7 As can be seen from (a), the THD of the A-phase current under the DPCC method at 40%, 60%, 80%, and 100% rated speeds are 9.80%, 9.74%, 8.40%, and 8.86%, respectively. The five-phase current fluctuation is significantly reduced compared to the LESO method, thus meeting the robustness requirements under parameter mismatch conditions. From Figure 7 As can be seen from (b), i corresponds to the DPCC speed. α The THDs of the currents were 9.80%, 9.45%, 8.24%, and 8.53%, respectively. Figure 7 As can be seen from (c), the fluctuation of the third harmonic current is around 0.59A, which is less than the amplitude of the third harmonic current of LESO.

[0189] from Figure 8 As can be seen from (a), the THD of the A-phase current under the MFAC method at 40%, 60%, 80%, and 100% rated speed are 6.01%, 3.71%, 3.18%, and 3.30%, respectively. The five-phase current fluctuation is significantly reduced compared to the LESO and DPCC methods, demonstrating better robustness. From Figure 8 As can be seen from (b), i corresponds to the MFAC speed. α The THDs of the currents were 5.70%, 3.38%, 2.67%, and 2.75%, respectively. Figure 8 As can be seen from (c), the fluctuation of the third harmonic current is around 0.59A, which is close to the amplitude of the third harmonic current of DPCC. In the middle part, MFACC is smaller.

[0190] 2. Variable speed operating conditions

[0191] The performance of the proposed method was evaluated under operating conditions of speeds from 400 rpm to 2000 rpm and a load torque set at 60% of the rated torque. The corresponding results are as follows: Figure 9-12 As shown.

[0192] from Figure 9 As can be seen in (a), the THD of the A-phase current under the LESO method at 40%, 60%, 80%, and 100% rated speed are 5.04%, 4.77%, 4.82%, and 4.76%, respectively. In the absence of parameter mismatch, the phase current fluctuation of the LESO method is significantly reduced. Figure 9 As can be seen from (b), i corresponds to the rotational speed of LESO. α The THDs of the currents were 4.45%, 4.18%, 4.08%, and 4.07%, respectively. Figure 9 As can be seen from (c), the harmonic current fluctuates by about 0.75A.

[0193] from Figure 10 As can be seen from (a), the THD of the A-phase current under the same rotational speed using the LESO method are 6.95%, 6.63%, 5.90%, and 5.64%, respectively, while the DPCC method exhibits more high-frequency harmonics. From Figure 10 As can be seen from (b), i corresponds to the rotational speed of LESO. α The THDs of the currents were 6.63%, 6.10%, 5.38%, and 5.08%, respectively. Figure 10 As can be seen from (c), the fluctuation of the third harmonic current is around 0.86A.

[0194] from Figure 11 As can be seen in (a), the THD of the A-phase current using the LESO method at the same speed is 4.73%, 4.84%, 5.90%, and 3.90%, respectively. Although the MFAC method appears smoother, it is evident that the THD is higher at 80% of the rated speed. Figure 11 As can be seen from (b), i corresponds to the MFAC speed. α The THDs of the currents were 3.03%, 3.16%, 5.20%, and 3.10%, respectively. Figure 11 As can be seen from (c), the fluctuation of the third harmonic current is around 0.72A.

[0195] 3. Variable torque operating condition

[0196] The aim was to verify the performance of the motor control algorithm under varying load conditions. During motor operation, the motor speed was fixed at 800 rpm, and the load torque was gradually increased from 40% to 100% of the rated torque to test the algorithm's performance. The results are as follows: Figures 12 to 14 As shown.

[0197] from Figure 12As can be seen from (a), the THD of the A-phase current under the LESO method at 40%, 60%, 80%, and 100% rated torque are 14.51%, 6.77%, 4.57%, and 3.44%, respectively. At low torque, the LESO method exhibits a relatively large THD. Figure 12 As can be seen from (b), i corresponds to the torque under LESO. α The THDs of the currents were 12.29%, 5.59%, 3.82%, and 2.92%, respectively. Figure 12 As can be seen from (c), the harmonic current fluctuates around 0.87A.

[0198] from Figure 13 As can be seen from (a), the THD of the A-phase current under the same torque using the DPCC method are 14.95%, 7.51%, 4.83%, and 3.35%, respectively. Compared with the LESO method, the THD of the DPCC method is not significantly different. Figure 13 As can be seen from (b), i corresponds to the torque under DPCC. α The THDs of the currents were 13.21%, 6.75%, 4.36%, and 2.85%, respectively. Figure 13 As can be seen from (c), the fluctuation of the third harmonic current is around 0.77A.

[0199] from Figure 14 As can be seen from (a), the THD of the A-phase current under the same torque using the MFACC method are 13.87%, 7.75%, 4.30%, and 4.60%, respectively. Compared with other methods, the MFACC method has a greater impact from low-order harmonics. Figure 14 As can be seen from (b), i corresponds to the torque under MFACC. α The THDs of the currents were 10.91%, 6.48%, 2.54%, and 4.22%, respectively. Figure 14 As can be seen from (c), the fluctuation of the third harmonic current is around 0.77A.

[0200] Experimental results show that the MFACC method exhibits performance close to that of the LESO and DPCC methods in both variable torque and variable speed experiments. However, in parameter mismatch experiments, the MFACC method outperforms the other two strategies.

[0201] In summary, this invention proposes a model-free adaptive current control (MFACC) parameter tuning method for five-phase permanent magnet synchronous motors (PMSMs) based on the center collision optimization algorithm (CCO). This method can dynamically linearize the d1q1 axis voltage and current sampling data of the PMSM without the physical parameters of the PMSM, and calculate the reference output voltage at the next moment using the d1q1 axis voltage and current sampling data. The optimal relevant parameters of the MFACC are optimized offline using the CCO optimization algorithm. NFV-SVPWM is used for PWM modulation in the five-phase PMSM, enabling the driving of the five-phase PMSM system with less sampling data. By comparing with the LESO and DPCC methods, the CCOFMACC method shows better robustness under parameter mismatch conditions. In variable speed and variable torque experiments, the performance of the three methods is similar. Therefore, the CCOFMACC method can ensure stable, fast, and robust control of the five-phase PMSM drive.

[0202] The technical means disclosed in this invention are not limited to those disclosed in the above embodiments, but also include technical solutions composed of any combination of the above technical features. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this invention, and these improvements and modifications are also considered within the scope of protection of this invention.

Claims

1. A method for tuning adaptive current control parameters of a permanent magnet synchronous motor, characterized in that, Includes the following steps: S1. Establish a mathematical model for a five-phase permanent magnet synchronous motor. Based on the topology of a five-phase permanent magnet synchronous motor driven by a two-level voltage source inverter, the phase bridge arm switching function is defined, 32 voltage vectors are generated and mapped to the α-β principal plane and the xy third harmonic plane. The third harmonic is eliminated by the nearest four vector space vector pulse width modulation (NFV-SVPWM) technology. The voltage transformation relationship between the stationary coordinate system and the synchronous rotating coordinate system is established, and the stator voltage equation, electromagnetic torque equation and mechanical motion equation of the d1-q1 axis and d3-q3 axis are obtained. S2. Design a model-free adaptive controller and a pseudo-block Jacobian matrix estimation algorithm. The five-phase permanent magnet synchronous motor system is represented as a multi-input multi-output nonlinear discrete-time system. Based on the full-format dynamic linearization FFDL, the nonlinear system is transformed into a linear data model. A model-free adaptive controller with cost function minimization is designed, and an estimation algorithm for the pseudo-block Jacobian matrix PPJM is constructed. Reset conditions are set to prevent parameter drift. S3. Offline tuning of model-free adaptive current control parameters based on center collision optimization algorithm The input and output variables of the model-free adaptive current controller are defined, the parameters to be optimized are determined, and an optimization function is established with the goal of suppressing the third harmonic current and reducing torque ripple. The center collision optimization algorithm is used to adaptively allocate the search ratio in the original space and the decorrelation space, and the control parameters are iteratively optimized. The optimal parameters are applied to the controller to realize the current control of the five-phase permanent magnet synchronous motor.

2. The method for tuning adaptive current control parameters of a permanent magnet synchronous motor according to claim 1, characterized in that: Step S1 includes: S11, Mathematical Model of Five-Phase Permanent Magnet Synchronous Motor In the topology of a five-phase permanent magnet synchronous motor (PMSM) system driven by a two-level voltage source inverter (2L-VSI), the phase voltage of each phase arm is directly determined by the switching state of the corresponding arm; the switching function of the p-th phase (p∈{a,b,c,d,e}) is defined as S. p ∈{0,1}: When the switch in the phase bridge arm is turned on, S p =1, when off, S p =0; Since there are two switching states for each phase arm, the 2L-VSI of the five-phase PMSM can generate 32 different combinations of switching signals; the mathematical expression of the phase voltage generated by these switching states is: (1) Among them, V p It is the stator phase voltage of phase p, V dc It is the DC bus voltage, S p This is the switching state of phase p; The 2L-VSI based on the five-phase PMSM generates 32 switching states, corresponding to 32 voltage vectors. These vectors are mapped to the α-β principal plane and the xy third harmonic plane, forming vector distributions with different amplitudes. Based on the amplitude, these vectors in the α-β plane are divided into 10 large vectors, 10 medium vectors, and 10 small vectors, respectively V l V m V s The corresponding voltage amplitudes are 0.6472Vdc, 0.4Vdc, and 0.2472Vdc, respectively, and there are two zero vectors; the voltage vectors in the α-β plane and the xy plane are denoted as V, respectively. αβ and V xy This voltage is obtained from the five-phase voltage transformation in the natural coordinate system and is expressed as: (2) Where j is the imaginary unit, and γ1 and γ3 are complex exponential rotation operators, corresponding to the phase offset coefficients of the fundamental and third harmonic components of the five-phase PMSM system, respectively. To eliminate the third harmonic in a five-phase permanent magnet synchronous motor, the nearest four-vector space vector pulse width modulation (NFV-SVPWM) technology is adopted. Two large vectors and two medium vectors are selected, and the third harmonic is eliminated by allocating appropriate execution times. The α-β subspace is divided into 10 sectors. In the first sector, the selected vectors are V25, V24, V16, V29, and zero vectors V0 and V31. Each voltage vector is synthesized into a voltage vector of 0 in the xy subspace according to the execution times T1, T2, T3, and T4. To achieve decoupled control of the stator voltage in the synchronous rotating coordinate system, the fundamental voltage vector V in the stationary coordinate system is... αβ and the third harmonic voltage vector V xy Transforming to the fundamental d1-q1 synchronous rotating coordinate system and the third harmonic d3-q3 synchronous rotating coordinate system respectively, the corresponding Park transformation relationships are expressed as follows: (3) Among them, V d1 V q1 V represents the d1-q1 axis voltage component in the fundamental subspace synchronous rotating coordinate system. d3 V q3 For the d3-q3 axis voltage components in the synchronous rotating coordinate system of the third harmonic subspace, θ e It is the electrical angle of the PMSM rotor; The stator voltage equations for the five-phase permanent magnet synchronous motor along the d1-q1 and d3-q3 axes are expressed as follows: (4) Among them, R s It is the stator resistance, i d1 i q1 i d3 i q3 These are the fundamental plane d1-q1 axis currents and the harmonic plane d3-q3 axis currents, respectively. d1 ,L q1 It is the stator inductance along the d1-q1 axis, L l It is the stator leakage inductance, D(·) is the time derivative, ω e It is the electric angular velocity of the rotor, ψ f It is the magnetic flux linkage of a permanent magnet; the electromagnetic torque of a five-phase PMSM is expressed as: (5) Among them, T e It is electromagnetic torque, P n This refers to the number of pole pairs of the motor; the mechanical motion equation of a five-phase PMSM is expressed as: (6) Where J is the moment of inertia, ω m It is the mechanical angular velocity of the rotor, T l It is the load torque, B m It is the coefficient of friction.

3. The method for tuning adaptive current control parameters of a permanent magnet synchronous motor according to claim 2, characterized in that: Step S2 includes: S21, FFDL of Multiple Input Multiple Output System The multi-input multi-output nonlinear discrete-time system is represented as: (7) Where u(k) and y(k) are the input and output data at time k in the multiple-input multiple-output system, and kn u It is the nth u Historical input data, kn y It is the nth y Historical output data for each; When designing a model-free adaptive current controller, a multi-input multi-output system must satisfy two assumptions: Assumption 1: Except for finite time points, nonlinear functions , i=1 to i=n y The partial derivatives of the +2 variables all exist; Assumption 2: Except at finite time points, the multiple-input multiple-output system satisfies the Lipschitz condition, that is, for any k1≠k2, k1,k2≥0 and u(k1)≠u(k2), it can be expressed as: (8) Where b is a constant greater than 0, L y and L u It is the length of the historical input / output data window, and satisfies 1≤L y ≤n y , 1≤L u ≤n u H Ly,Lu (k) is a length of L y +L u The information vector is represented as: (9) The system satisfies the above two assumptions, for any fixed L y and L u , ||ΔH Ly,Lu If (k)||≠0, then there exists a time-varying pseudo-block Jacobian matrix PPJM as Φ f,Ly,Lu (k) transforms the nonlinear system into a fully dynamically linearized FFDL data model, represented as: (10) Where, ΔH Ly,Lu (k)=H Ly,Lu (k)-H Ly,Lu (k-1) is the difference between the elements in the information vector, Φ f,Ly,Lu (k)=[φ1(k),…,φ Ly (k),φ Ly +1(k),…,φ Ly+Lu [(k)] is a PPJM matrix, and ||Φ f,Ly,Lu (k)|| is a bounded matrix, represented as: (11) Where b1 ≠ 0, b2 > 0, a > 1, m is the dimension of the input and output, satisfying b2 > b1(2a+1)(m-1), φ ij(Ly+1) It is an element in the (Ly+1)th square block; S22, Model-Free Adaptive Controller Design Based on the full-format dynamic linearized FFDL data model (10), the cost function of the controller is expressed as: (12) Among them, y * (k+1) is the desired output of the controller, and λ is a weighting factor. The nonlinear discrete-time system (7) is introduced into the cost function (12) so that the derivative of the cost function (12) with respect to u(k) is equal to 0, thus obtaining the minimum value of the cost function (12). Therefore, the controller is expressed as: (13) Where Δy(k) = y(k) - y(k-1) and Δu(k) = u(k) - u(k-1) are the differences between the input and output, and ρ i i=1,2, …,L y +L u It is L y +L u A pseudo-order weighting factor; S23, PPJM estimation algorithm Based on the full-format dynamic linearized FFDL data model (10), the cost function for pseudo-block Jacobian matrix estimation is expressed as: (14) in, yes The estimated value, μ is a weighting factor that makes the cost function (14) effective for... The derivative of is equal to 0, which gives the minimum value of the cost function (14). Therefore, the pseudo-block Jacobian matrix estimate is expressed as: (15) To prevent parameter drift and meaningless estimation, it is necessary to... Set the reset conditions as follows: (16) (17) Where a > 1, m is the system input / output dimension, η is the step size factor, μ is the weight factor, and b1 and b2 are two integers. yes The initial value.

4. The method for tuning adaptive current control parameters of a permanent magnet synchronous motor according to claim 3, characterized in that: Step S3 includes: S31. Parameter optimization of model-free adaptive current control When applying model-free adaptive current control (MFACC) to the current control of a five-phase PMSM mathematical model, preliminary parameter tuning of MFACC is required. First, the inputs and outputs y(k) and u(k) of the MFACC method are defined as follows: (18) Among them, i d1 (k), i q1 (k) and u d1 (k), u q1 (k) represents the historical current and voltage input data of the MFACC controller, defined as u d1 (k+1), u q1 (k+1) is the voltage output result of the MFACC controller; The input and output dimensions of the driving system are m=2. In the pseudo-block Jacobian matrix, weight parameters η and μ are selected, with a lower bound of b²=1e-6 and an upper bound of a*b²=50. The matrix elements are initialized. φ 11 , φ 22 , φ 12 and φ 21 Not setting b1 can reduce the number of elements φ. 12 and φ 21 Impact; Selecting the length L of historical data y =3 and L u =3; Select parameter ρ in controller u(k). i and λ, ρ i We use a single parameter ρ to replace the others; therefore, the center collision optimization algorithm needs to optimize 6 parameters, namely... ; Select a set of optimization parameters Using [0.3, 80, -8, 3, -0.5, -5] as a baseline, these parameters are adjusted up and down relative to their rated values; since η and μ have little impact on drive performance, the analysis focuses on [ρ, λ, φ 11 , φ 22 The effects of ]; S32. Optimize the objective function Suppressing harmonic currents and reducing torque ripple are important performance indicators for five-phase PMSM current control. When using the model reference adaptive control method under various operating conditions, uppercase J is selected as the optimization objective to quantify the control performance of the five-phase PMSM. (19) In the formula, N is the total number of sampling points in the motor control process, and λ J is the weighting coefficient of the third harmonic current, k is the current sampling point, and k∈[1,N]; the optimization objective J is constructed by comparing the actual d1-q1 axis current sampling value at the kth sampling point with its reference value, aiming to achieve real-time tracking of the actual d1-q1 axis current to the reference value and reduce the amplitude of the actual d3-q3 axis current; S33. Offline MFACC parameter tuning method Step 1: Collect historical dq-axis voltage and current data of the five-phase PMSM in a synchronous rotating coordinate system, construct a data-driven model based on FFDL, design the MFACC controller, and establish an estimation algorithm for the PPJM; determine the lengths Ly and Lu of the historical data, and select the parameters of the MFACC controller. ; Step 2: Set the experimental operating conditions for the five-phase PMSM; integrate the MFACC controller into the drive system of the five-phase PMSM, and set the parameters to be optimized. Integrate it into the CCO algorithm so that it can be called offline; Step 3: Configure the initialization process of the CCO algorithm; determine the parameters. Given the constraints of the boundary conditions, design the objective optimization function J; Step 4: Employ the CCO optimization algorithm to iteratively optimize the set of control parameters in step S3 offline, minimizing the objective function J, and outputting the optimal parameters. ; Step 5: Apply the optimized parameters to the MFACC controller to achieve current control of the five-phase PMSM.

Images