Permanent magnet synchronous motor model-free predictive current control method with vector preselection

Abstract

This invention provides a model-free predictive current control (MFPCC) method for permanent magnet synchronous motors (PMSMs) with vector preselection, based on a hyperlocal model. Based on the hyperlocal model, this invention proposes a low-complexity multi-vector MFPCC. First, a hyperlocal model is used to replace the motor model in traditional MPC, reducing the impact of motor parameter mismatch. Then, to estimate the disturbance term in the hyperlocal model, a sliding diaphragm observer based on a variable-gain reaching law is designed. Finally, a three-vector algorithm with vector preselection is proposed. This preselection scheme evaluates the vector position based on the voltage gradient concept, directly obtaining the optimal voltage vector without traversing vectors and evaluating cost functions, thus reducing computation time.

Description

Technical Field

[0001] This invention relates to the field of power electronics, and specifically to a model-free predictive current control method for permanent magnet synchronous motors with vector preselection based on a hyperlocal model. Background Technology

[0002] Permanent magnet synchronous motors (PMSMs) are increasingly in demand across various fields due to their compact size, high power density, excellent reliability, and wide speed range, such as new energy vehicles, high-end CNC systems, and rail traction systems. Classic control strategies for PMSMs include FOC (Functional Magnet Control) and DTC (Distributed Torque Control). However, FOCs with proportional-integral controllers generally require a suitable inner-loop bandwidth to avoid system instability and output oscillations. Furthermore, the inner-loop bandwidth may limit the dynamic performance of the speed loop. Compared to FOCs, DTCs, with their simpler structure, achieve better torque control performance. On the other hand, DTCs exhibit drawbacks at low speeds, including low controllability, high torque ripple, variable switching frequency, and high noise.

[0003] In recent years, FCS-MPC has attracted much attention from scholars as a powerful algorithm due to its intuitive concept, real-time optimization characteristics, multi-objective control, and convenient constraint optimization. However, the MPC algorithm applies only one voltage vector with a fixed amplitude and phase in each control cycle, which makes it difficult to reduce fluctuations in current and torque of the PMSM. Furthermore, due to the current prediction and iterative optimization process in each cycle, MPC faces a heavy computational burden. This computational cost is even greater in multi-vector algorithms.

[0004] To address the aforementioned issues, some scholars have proposed low-complexity multi-vector algorithms. For example, some have proposed using virtual voltage vectors to expand the candidate set of vectors, and then limiting the number of vectors by establishing a torque and flux linkage lookup table (LUT); others have directly obtained the optimal voltage vector based on the position of the reference voltage vector, reducing the number of candidate voltage vectors from 6 to 1. Valid and zero voltage vectors are included in the suboptimal voltage vector candidate set to effectively improve torque ripple. However, this method involves complex arctangent operations and still consumes certain computational resources; still others have proposed a voltage pre-selection scheme based on the motor rotation direction and the previous optimal voltage vector, which limits the number of candidate vectors to 3.

[0005] On the other hand, FCS-MPC is highly dependent on an accurate motor model. To enhance the robustness of the system, some researchers have studied model-free predictive current control (MFPCC) based on hyperlocal models. Some researchers have used extended state observers to estimate the disturbance term of the ULM. Furthermore, they analyzed the effect of the model-free algorithm at different sampling frequencies, demonstrating that this method improves the steady-state and dynamic performance at low sampling frequencies and simplifies the debugging process. Other researchers have considered the mismatch disturbances in the motor control system, including external loads and mechanical parameter disturbances, and proposed a model-free algorithm with nonlinear disturbance compensation. Some researchers have designed an MFPCC based on the constant-speed sliding mode approaching law (CRRL-MFPCC). Through the sliding mode observer (SMO), the disturbance term in the ULM can be estimated, thus eliminating the stagnation problem in current change updates. Furthermore, an effective voltage vector and a zero vector are applied in one control cycle, effectively improving the current tracking performance of the system. However, because this method uses a constant-rate approach law, it is difficult to balance response speed and chattering level. Duty cycle modulation schemes can only synthesize voltage vectors with a fixed phase, limiting the system's control performance.

[0006] Therefore, it is of great research value to design a suitable sliding diaphragm reaching law for ULM-based MFPCC while considering issues such as multi-vector modulation and reducing computational cost.

[0007] The drawbacks of existing technologies are as follows: Traditional MPC operates on only one voltage vector per switching cycle, resulting in significant current and torque fluctuations in PMSMs, and placing a heavy computational burden on MPCs. This computational cost is even greater in multi-vector algorithms. Furthermore, traditional MPCs are highly dependent on accurate motor models, making them unsuitable for applications where accurate models are difficult to establish. Additionally, in ULM-based MFPCCs, the sliding diaphragm observer typically uses a constant-velocity reaching law, which may cause chattering. Summary of the Invention

[0008] The objective of this invention is achieved through the following technical solutions.

[0009] This invention proposes a low-complexity multi-vector MFPCC based on ULM. First, ULM is used to replace the motor model in traditional MPC, reducing the impact of motor parameter mismatch. Then, to estimate the perturbation term in ULM, a sliding diaphragm observer based on a variable gain reaching law is designed. Finally, a three-vector algorithm with vector pre-selection is proposed. This pre-selection scheme evaluates vector positions based on the voltage gradient concept, directly obtaining the optimal voltage vector without traversing vectors or evaluating cost functions, thus reducing computation time.

[0010] Specifically, this invention provides a model-free predictive current control method for permanent magnet synchronous motors with vector preselection based on a hyperlocal model, comprising: Step 1: Establish a hyperlocal model of the permanent magnet synchronous motor, wherein the hyperlocal model includes parameter perturbation terms for the dq axis; Step 2: Establish a sliding mode observer based on the variable gain reaching law to estimate the parameter perturbation term in the hyperlocal model. The sliding mode gain of the variable gain reaching law changes with the sliding surface, which suppresses chattering while ensuring rapid reaching. Step 3: Calculate the reference voltage and perform voltage vector pre-selection. Calculate the reference voltage vector based on the current reference value and the parameter disturbance term estimated in Step 2. Then, based on the voltage gradient, directly select the first and second optimal voltage vectors in a non-ergodic manner. Step 4: Duty cycle calculation and vector synthesis. Based on the two optimal voltage vectors selected in Step 3, a zero vector is introduced to form a three-vector modulation scheme. The duty cycle of each vector is calculated and the final voltage vector is synthesized to drive the permanent magnet synchronous motor.

[0011] The advantages of this invention are: 1. The ULM-based three-vector MFPCC proposed in this invention has better steady-state performance and can effectively reduce the system's sensitivity to parameters; 2. The sliding diaphragm observer based on the variable gain reaching law proposed in this invention is used to obtain the perturbation term of the ULM. The sliding diaphragm gain changes with the state variables of the system, which effectively suppresses the chattering of the observer. 3. Considering the complexity of multi-vector algorithms, the vector pre-selection scheme proposed in this invention directly selects the first optimal voltage vector using the gradient of the reference voltage vector. Based on the designed pre-selection rules, the second optimal voltage vector can be obtained without current prediction and cost evaluation processes, greatly reducing the computational burden. Attached Figure Description

[0012] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A block diagram of a conventional MPCC control is shown.

[0013] Figure 2 A voltage vector distribution diagram of a conventional MPCC is shown.

[0014] Figure 3 A newly divided sector distribution map according to an embodiment of the present invention is shown.

[0015] Figure 4 An embodiment of the present invention is shown. The diagram illustrates the pre-selection rules for this event.

[0016] Figure 5 An embodiment of the present invention is shown. A diagram illustrating the pre-selection rules at that time.

[0017] Figure 6 A control block diagram of the proposed MFPCC according to an embodiment of the present invention is shown.

[0018] Figure 7 A schematic diagram of an electric motor experimental platform according to an embodiment of the present invention is shown.

[0019] Figure 8 The steady-state experimental waveform under experimental condition 1 is shown.

[0020] Figure 9 The steady-state experimental waveform under experimental condition 2 is shown.

[0021] Figure 10 The steady-state experimental waveform under experimental condition 3 is shown.

[0022] Figure 11 The waveform diagram of the parameter mismatch experiment under experimental condition 4 is shown.

[0023] Figure 12 The waveform diagram of the parameter mismatch experiment under experimental condition 5 is shown. Detailed Implementation

[0024] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0025] Terminology Explanation: Finite Control Set Model Predictive Control (FCS-MPC): FCS-MPC first establishes a discrete mathematical model of the system, then predicts the system state at future times based on a finite set of inputs. It then uses rolling optimization to calculate the control input that minimizes the cost function and applies it to the next control cycle. FCS-MPC can fully consider the nonlinear characteristics of the system and handle multi-objective constraints. However, the prediction process and cost function optimization consume significant computational resources, placing high demands on processor performance. Furthermore, FCS-MPC is a model-dependent control method; mismatch in model parameters can degrade the system's control performance.

[0026] Model-Free Predictive Control (MFPC): MFPC is a predictive control method that does not rely on a system model. It constructs an equivalent model online using only the system's input and output data to predict the system's state at future times. This method has advantages such as strong adaptability to nonlinear systems, low parameter sensitivity, and low modeling burden, making it suitable for power electronic systems where it is difficult to establish accurate models.

[0027] Ultra-Local Model (ULM): ULM is a commonly used simplified model in MFPC (Multi-Functional Control Programming) to replace the actual model of the control system. By estimating the disturbance terms of the ULM using a sliding mode observer (SMO), and then discretizing the model, prediction and optimization can be achieved without building an accurate model. However, the sliding mode observer may produce chattering, requiring the design of a sliding mode reaching law to ensure appropriate reaching rates and chattering levels.

[0028] A. System Model The controlled object of this invention is a surface-mounted permanent magnet synchronous motor, whose mathematical model in a rotating coordinate system can be written as: In the formula, and These represent the stator voltages along the d and q axes, respectively. and These represent the stator currents along the d and q axes, respectively. Indicates stator resistance; Indicates stator resistance; Represents electric angular velocity; This indicates the magnetic flux linkage of a permanent magnet.

[0029] In the model-free strategy proposed in this invention, the current prediction is calculated using a discretized hyperlocal model. The first-order hyperlocal model of the single-input single-output system is expressed as: In the formula, and These represent the system's input and output, respectively. It is an uncertainty term of the system; Represents the gain factor.

[0030] To eliminate the influence of motor parameter mismatch, a hyperlocal model is introduced into the motor model. According to (1), the hyperlocal model of the motor can be reconstructed as: In the formula, , and These represent the parameter perturbation terms for the dq axis, respectively.

[0031] B. Traditional MPCC Algorithm Based on (1), according to the first-order forward Euler formula, the current prediction model in the discrete domain can be derived as follows: In the formula, and Let represent the d-axis voltage and current at time k, respectively; and Let represent the q-axis voltage and current at time k, respectively; and The predicted value of the dq-axis current at time k+1; Indicates the control cycle.

[0032] The control block diagram of the traditional MPCC algorithm is as follows: Figure 1 As shown. A two-level voltage source inverter has eight switching states, providing eight voltage vectors. The voltage vector distribution diagram is shown below. Figure 2 As shown. In the discretized motor model, the predicted current value can be calculated by applying all voltage vectors. However, in practical applications, due to the existence of digital delays such as dead time and filtering delay, first-order delay compensation needs to be considered in the current prediction process. Therefore, (4) should be reconstructed as follows: In the formula, and This represents the predicted current value at time k+2; and This represents the candidate voltage vector at time k+1.

[0033] Based on the principle of minimizing current deviation, the voltage vector that minimizes the cost function value is selected as the optimal voltage vector. The cost function of the traditional MPCC algorithm is designed as follows: As discussed above regarding the traditional MPC algorithm, it involves complex current prediction and vector optimization processes, leading to a significant computational burden. Furthermore, the traditional MPC algorithm applies only one voltage vector per control cycle, resulting in substantial current and torque ripples. While multi-vector modulation schemes are considered an effective solution, the calculation of vector application time further consumes computational resources. Additionally, motor parameters are involved in the mathematical calculations of MPC. Inaccurate motor parameters have a very negative impact on the selection of the optimal voltage vector, with inductance and flux linkage parameters having the most significant effect.

[0034] C. Traditional SMO To enhance the robustness of the motor model, a sliding diaphragm observer based on the constant velocity reaching law was used to obtain the perturbation term. and From the perspective of the hyperlocal model of the motor, this sliding diaphragm observer can be constructed as follows: In the formula, and These are the observed values ​​of stator current and uncertain disturbances; These are synovial parameters; This represents the synovial control function.

[0035] Typically, the sliding surface is designed such that the difference between the observed and actual current values ​​is: The synovial convergence law is: In the formula, s represents the sliding surface; k is the sliding gain; It is a symbolic function.

[0036] Combining (7)-(9), the sliding diaphragm observer can easily acquire the disturbance term due to the use of the constant-speed approach law. The sliding diaphragm gain k determines the approach speed of the observer; however, when the gain is increased to speed up the response, the chattering of the system also increases; conversely, reducing the gain sacrifices the response speed to ensure a smaller chattering level. D. The SMO proposed in this invention According to (7), in order to estimate the dq-axis current and disturbance, the sliding observation in the rotating coordinate system can be rewritten as: In the formula, and These are the observed values ​​of the dq-axis currents, respectively. and These are the observed values ​​of the dq-axis perturbation; and Indicates synovial parameters; and It is the dq-axis sliding control function. According to (3), (10) and (11), the corresponding dq axis difference equation can be expressed as: In the formula, and It is the difference between the observed and actual values ​​of the dq-axis current; and It is the difference between the observed and actual values ​​of the dq-axis perturbation term.

[0037] The core of sliding mode theory lies in using a sliding mode control function to dynamically approach a low-order sliding surface. In the model, the dq-axis linear sliding surface is: Traditional sliding diaphragm reaching laws have a fixed gain, making it difficult to simultaneously meet the requirements of high response speed and low chattering level. Therefore, this invention uses a reaching law with a variable gain term. This reaching law is expressed as: In the formula, , , , , As the system state approaches the sliding surface, the variable gain term G gradually approaches the equilibrium state. Conversely, when the system state is far from the sliding surface, the variable gain term G approaches a constant. And because of this, the gain is greater than Therefore, compared to the constant-rate reaching law, this reaching law improves the convergence speed of the system to some extent. In general, as the system state approaches the sliding surface, the gain term G changes with the sliding surface... The gain changes over time. Approximately as it approaches the equilibrium point, the gain becomes closer to... This allows for the suppression of chattering levels through synovial control.

[0038] According to (12), (13) and (14), (15) can be described as: and The disturbances considered to be in the control function are designed as follows: To ensure the stability of the sluice box observer, its parameters must be carefully chosen. The Lyapunov stability criterion is used to prove the stability of this reaching law; therefore, the observer must satisfy the following condition: For ease of calculation, let's record... Substituting (12) and (17) into (18) yields the stability equation: Since the value of M is in In order to satisfy the stability condition (18), the parameters are within the interval. and The following expression must be satisfied: Similarly, the parameters of the q-axis can be obtained. and The following conditions must be met: From (20) and (21), it can be seen that in order to obtain an overall stable effect, the final parameters should satisfy: Based on the stability analysis above, a sliding mode observer with appropriate parameters can reach the sliding surface within a finite time and make the current and disturbance deviations approach 0.

[0039] In practical applications, the synovial observer needs to be discretized based on the Euler equations. According to (10) and (11), the discretized observer model is derived as follows: in, and The observed value of the predicted dq-axis current at time k+1; and These are the observed values ​​of the dq-axis current at time k; and These are the dq-axis sliding control functions at time k.

[0040] E. Vector Preselection and Modulation Scheme To enhance current control performance, a three-vector algorithm is introduced into model-free control. However, the increase in vector combinations complicates the optimization and modulation processes, resulting in a high computational burden. Therefore, a voltage gradient-based vector pre-selection scheme is proposed to simplify the algorithm's complexity.

[0041] To avoid the calculation of the arctangent function, this scheme re-divided the sectors, creating six new sectors. For example... Figure 3 As shown, each basic voltage vector is located at the center of a sector, and each sector is bounded by two straight lines with gradients. When the reference voltage vector is located in a certain sector, the system determines... The range of the reference voltage gradient in the plane makes it easy to determine the sector number, and thus directly select the first optimal voltage vector. The dq-axis reference voltage equation is expressed as: Using the inverse Park transform The shaft reference voltage can be obtained as follows: From the above analysis, the gradient of the reference voltage vector is: When the reference voltage vector is located in sector I, the voltage gradient is greater than the line. The gradient is less than The gradient. Therefore, the range of the voltage gradient in sector I can be expressed as: Similarly, the gradient ranges for other sectors can also be derived. Therefore, based on the gradients of the sector boundary lines, a lookup table can be created for vector pre-selection. This can be seen from Table 1. The selection of this method omits the current prediction and cost function evaluation processes. Furthermore, this method avoids the use of the arctangent function for position calculation during sector determination.

[0042] Table 1. Selection of the first optimal voltage vector Selection of the second optimal voltage vector and reference voltage vector and the first optimal voltage vector It is related to the gradient. For example, Figure 4 As shown, when In the upper half of sector I, yes and ( The vector difference is shown in the figure. The gradient is less than The gradient and the second optimal voltage vector Selected as .when In the lower half of sector I, it can be seen that... The gradient is greater than The gradient. It was selected as the second optimal voltage vector. In sector I, this conclusion can be expressed as: in and These represent the subscripts of the first and second optimal vectors, respectively; yes and The gradient of the vector difference; yes The gradient. Similarly, the other 5 sectors exhibit the same pattern. The prerequisite for this rule to hold is: in and Representing the reference and the first optimal vector respectively Axial components.

[0043] When equation (30) is not true, such as Figure 5 As shown, the four regions marked by dashed lines contain... Taking region A as an example, Selected as , The gradient is less than 0. Selected as When the reference voltage vector Located in region B, Selected as , The gradient is greater than 0. Selected as Two opposing regions exhibit the same pattern; that is, regions A and C follow the same pattern, and regions B and D follow the same pattern. This pattern can be represented as: Therefore, according to (29)-(31), the vector preselection scheme proposed in this invention does not require current prediction and cost evaluation, which greatly reduces computational resources. Based on the voltage gradient, the scheme can select the first and second optimal voltage vectors through a simple judgment process.

[0044] To reduce the computation time of multi-vector modulation, the proposed solution uses a cost function to calculate the duty cycle of the optimal vector. , and zero vector ( The duty cycles of ) are respectively set to , , The cost function for the three voltage vectors: In the formula, and They are respectively of Axis cost function and voltage; and They are respectively of Axis cost function and voltage; and They are respectively of Axis cost function and voltage, .

[0045] Based on the deadbeat principle, the duty cycle calculation equation is designed as follows: , and The duty cycle is solved as: Finally, after vector pre-selection and duty cycle calculation, the optimal voltage vector is synthesized to enhance current tracking performance and reduce current harmonics. The control block diagram of the MFPCC strategy proposed in this invention is shown below. Figure 6 As shown, it includes: Step 1: Establish a hyperlocal model of the permanent magnet synchronous motor, wherein the hyperlocal model includes parameter perturbation terms for the dq axis; Step 2: Establish a sliding mode observer based on the variable gain reaching law to estimate the parameter perturbation term in the hyperlocal model. The sliding mode gain of the variable gain reaching law changes with the sliding surface, which suppresses chattering while ensuring rapid reaching. Step 3: Calculate the reference voltage and perform voltage vector pre-selection. Calculate the reference voltage vector based on the current reference value and the parameter disturbance term estimated in Step 2. Then, based on the voltage gradient, directly select the first and second optimal voltage vectors in a non-ergodic manner. Step 4: Duty cycle calculation and vector synthesis. Based on the two optimal voltage vectors selected in Step 3, a zero vector is introduced to form a three-vector modulation scheme. The duty cycle of each vector is calculated and the final voltage vector is synthesized to drive the permanent magnet synchronous motor.

[0046] To demonstrate the effectiveness of the proposed method, the control algorithm was verified on an electronic control platform based on TMS320F28335. Figure 7 The experimental platform used is shown. The parameters of the motor are described in Table 2. In the experiment, the sampling frequency and control frequency were set to 10 kHz. To experimentally demonstrate the advantages of the proposed strategy, conventional MPCC (CMPCC) and MFPCC based on constant-speed sliding diaphragm reaching law (CRRL-MFPCC) were used for comparative verification.

[0047] Table 2. Motor Parameters The experimental conditions are set as follows: Experimental conditions 1: The motor speed is set to 1000 r / min and the load is 1 Nm.

[0048] Experimental Condition 2: The motor speed was set to increase from 500 r / min to 1000 r / min, and the load was 1 Nm.

[0049] Experimental conditions 3: The motor speed was set to 500 r / min, and the load was increased from 0.5 Nm to 1 Nm.

[0050] Experimental Condition 4: The motor speed was set to 1000 r / min, the load to 0.5 Nm, and the motor parameters were changed to 0.5 times the nominal parameters (0.5). 0.5 0.5 ).

[0051] Experimental Condition 5: The motor speed is set to 1000 r / min, the load is 0.5 Nm, and the motor parameters are changed to 1.5 times the nominal parameters (1.5). 1.5 1.5 ).

[0052] Experimental effects of the present invention A. Steady-state performance evaluation Table 3. Current performance evaluation under experimental condition 1 Figure 8 The waveforms of motor speed, A-phase current, and dq-axis current are described. To analyze the system's steady-state performance in depth, the standard deviation is used to assess the fluctuation of the dq-axis current. , ).

[0053] Figure 8 Steady-state experimental waveforms under experimental condition 1. (a) CMPCC; (b) CRRL-MFPCC; (c) MFPCC proposed in this invention.

[0054] Table 3 shows that the dq-axis current ripple of CMPCC is the largest among the three methods, which is attributed to the application of a single vector in each control cycle. The dq-axis ripples of the proposed strategy are 0.0388 and 0.0218, respectively, further demonstrating the superior current control performance under multi-vector modulation. Compared to CMPCC, the other two strategies exhibit lower THD, with the CRRL-MFPCC strategy having a THD of 8.59% and the proposed strategy having a THD of 2.87%. Figure 8 As can be seen from (c), the proposed strategy has a clear advantage in both THD value and current ripple.

[0055] B. Dynamic performance evaluation Figure 9 Dynamic experimental waveforms under experimental condition 2. (a) CMPCC; (b) CRRL-MFPCC; (c) The proposed MFPCC. Figure 10 Dynamic experimental waveforms under experimental condition 3. (a) CMPCC; (b) CRRL-MFPCC; (c) The proposed MFPCC.

[0056] Figure 9 The waveforms of motor speed, A-phase current, and dq-axis current under the condition of sudden speed increase are described. Figure 10 The waveforms of motor speed, A-phase current, and dq-axis current under torque variations are described. From these waveforms, it can be seen that the speed and dq-axis current of both control strategies can reach the reference values ​​relatively quickly and stably track them. Figure 9 It is known that the response time of the three strategies is approximately 0.26 seconds. Similarly, as... Figure 10As shown, the CMPCC exhibits the fastest current response when the load torque suddenly increases to 1 Nm. Compared to the CRRL-MFPCC, although their dynamic response speeds are similar, the proposed strategy results in lower current ripple.

[0057] C. Parameter mismatch experiment Figure 11 : Experimental waveforms of parameter mismatch under experimental condition 4. (a) CMPCC; (b) CRRL-MFPCC; (c) The proposed MFPCC. Figure 12 : Parameter mismatch experimental waveforms under experimental condition 5. (a) CMPCC; (b) CRRL-MFPCC; (c) The proposed MFPCC.

[0058] To demonstrate the robustness of the proposed strategy, the performance of the three methods is evaluated under the condition that the motor inductance flux linkage parameters are changed simultaneously. Figure 11 and 12 The waveforms of the A-phase and dq-axis currents are described. From... Figure 12 In Figure (a), a significant deviation of the actual dq-axis current of the CMPCC strategy from the reference value can be observed. This indicates that inaccurate motor parameters lead to prediction bias in the traditional MPC. To obtain more intuitive experimental data, Table 4 shows the current performance indicators. Clearly, compared to CMPCC, CRRL-MFPCC and the proposed MFPCC have lower ripple and THD, which verifies the positive effect of the hyperlocal model in reducing parameter mismatch. As shown in Table 4, even when the motor parameters are changed by 0.5 or 1.5 times, the strategy proposed in this invention still maintains superior current performance and exhibits strong robustness.

[0059] Table 4. Current performance evaluation under parameter mismatch conditions D. Calculation time The actual execution time of each part of the algorithm is recorded in Table 5. CMPCC requires 12.52 μs of computation time, while CRRL-MFPCC occupies 18.53 μs in each control cycle. The proposed strategy has an execution time of 12.56 μs, with voltage vector preselection consuming only 0.72 μs. Therefore, although the strategy proposed in this invention includes multi-vector modulation and a sliding diaphragm observer, its computation time is still similar to that of the CMPCC algorithm, which proves that the preselection and modulation schemes effectively reduce the complexity of the algorithm.

[0060] Table 5. Calculation Time The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A model-free predictive current control method for permanent magnet synchronous motors with vector preselection, based on a hyperlocal model, characterized in that, include: Step 1: Establish a hyperlocal model of the permanent magnet synchronous motor, wherein the hyperlocal model includes parameter perturbation terms for the dq axis; Step 2: Establish a sliding mode observer based on the variable gain reaching law to estimate the parameter perturbation term in the hyperlocal model. The sliding mode gain of the variable gain reaching law changes with the sliding surface, which suppresses chattering while ensuring rapid reaching. Step 3: Calculate the reference voltage and perform voltage vector pre-selection. Calculate the reference voltage vector based on the current reference value and the parameter disturbance term estimated in Step 2. Then, based on the voltage gradient, directly select the first and second optimal voltage vectors in a non-ergodic manner. Step 4: Duty cycle calculation and vector synthesis. Based on the two optimal voltage vectors selected in Step 3, a zero vector is introduced to form a three-vector modulation scheme. The duty cycle of each vector is calculated and the final voltage vector is synthesized to drive the permanent magnet synchronous motor.

2. The model-free predictive current control method for a permanent magnet synchronous motor with vector preselection according to claim 1, characterized in that, The variable gain reaching law in step 2 is specifically as follows: This convergence law is expressed as: In the formula, , , , , ; s represents the sliding surface; k is the sliding gain; t represents time; It is a symbolic function.

3. The model-free predictive current control method for a permanent magnet synchronous motor with vector preselection according to claim 1, characterized in that, The stability of the described reaching law is proved using the Lyapunov stability criterion.

4. The model-free predictive current control method for a permanent magnet synchronous motor with vector preselection according to claim 1, characterized in that, The voltage vector pre-selection in step 3 specifically includes: Sector re-division: The voltage vector plane is re-divided into six new sectors with the gradient line of the basic voltage vector as the boundary; First optimal voltage vector selection: Calculate the gradient of the reference voltage vector, and then determine... The range of the reference voltage gradient under the plane determines the new sector to which it belongs, and the basic voltage vector at the center of the new sector is the first optimal voltage vector. Second optimal voltage vector selection: Compare the magnitudes of the reference voltage vector gradient and the first optimal voltage vector gradient, and combine the component relationships on the αβ axis to directly determine the second optimal voltage vector according to preset rules.

5. A model-free predictive current control method for a permanent magnet synchronous motor with vector preselection according to claim 1, characterized in that, The duty cycle calculation in step 4 is based on the no-difference principle, and the duty cycle of the optimal vector is calculated using a cost function.

6. The model-free predictive current control method for a permanent magnet synchronous motor with vector preselection according to claim 1, characterized in that, The hyperlocal model is: In the formula, and These represent the stator voltages along the d and q axes, respectively. and These represent the stator currents along the d and q axes, respectively. , and These represent the parameter perturbation terms along the d and q axes, respectively. Indicates stator resistance; Indicates stator resistance; Represents electric angular velocity; This indicates the magnetic flux linkage of a permanent magnet.

Images