Permanent magnet synchronous motor parameter estimation and robust control method

Abstract

The application relates to a permanent magnet synchronous motor parameter estimation and robust control method, relating to the technical field of motor control, which comprises the following steps: determining a continuous global feasible region according to the physical limit of a to-be-estimated parameter and a closed-loop control system stability criterion, and discretizing the continuous global feasible region into a finite candidate set; predicting current values corresponding to each candidate parameter based on a discrete current prediction model containing the to-be-estimated parameter; constructing a value function with actual sampling current as a reference benchmark, and screening an optimal estimation parameter corresponding to a minimum value function value through traversal optimization; and substituting the optimal estimation parameter into a voltage equation of a dead-time-free prediction current control to calculate a reference voltage vector, so that closed-loop robust control is realized. The application does not need to set the observer gain, adaptive law coefficient and the like, effectively compensates for current prediction deviation caused by parameter mismatch, significantly improves the robustness of dead-time-free prediction current control under parameter uncertainty, and reduces the engineering implementation difficulty.

Description

Technical Field

[0001] This invention relates to the field of motor control technology, and in particular to a method for parameter estimation and robust control of a permanent magnet synchronous motor. Background Technology

[0002] Permanent magnet synchronous motors (PMSMs) are widely used in new energy power generation, aerospace, and rail transportation due to their high power density and efficiency. Deadbeat predictive current control (DMC) has become one of the mainstream solutions for current control of high-performance PMSMs due to its fast dynamic response. However, the performance of DMC is highly dependent on an accurate mathematical model of the motor. In actual operation, motor parameters can be perturbed by factors such as temperature rise, magnetic saturation, and aging, leading to a mismatch between the nominal parameters used in the controller and the actual motor parameters. This results in problems such as inaccurate current prediction, increased current ripple, aggravated torque pulsation, and steady-state error, severely deteriorating the system's control performance. Therefore, accurate online estimation of key motor parameters is crucial for improving the robustness of DMC.

[0003] To address the aforementioned parameter mismatch problem, scholars both domestically and internationally have conducted extensive research. Existing online state estimation schemes can be broadly categorized into two main technical routes based on their core solution approaches: observer estimation and adaptive iterative identification. Observer estimation methods unify parameter perturbations, inverter nonlinearities, and external disturbances into a lumped disturbance term. Real-time estimation and feedforward compensation of the lumped disturbance are achieved by constructing a closed-loop state observer. These methods mainly include sliding mode observers, Kalman filters, extended state observers, and multi-element composite disturbance observers. Among these, sliding mode observers suppress parameter disturbances by constructing a sliding surface and designing a corresponding approaching control law. However, the sliding mode gain, as a core parameter, has a higher suppression capability for parameter disturbances but exacerbates inherent system chattering. Therefore, fine-tuning of the sliding mode gain is necessary to achieve a performance trade-off. Kalman filters require pre-setting the process noise covariance matrix and the measurement noise covariance matrix and involve multi-order matrix inversion and multiplication operations, resulting in high computational overhead and demanding high hardware computing power. While extended state observers can estimate disturbances in real time and reduce state observation errors, the adaptive law gain in their linear observation error equation requires repeated tuning, which is detrimental to engineering applications and widespread adoption. Adaptive iterative identification methods rely on the discrete voltage and current mathematical model of the motor to construct parameter update equations, iteratively updating parameter estimates by minimizing the prediction error value function. These methods mainly include recursive least squares, signal injection, model reference adaptive systems, and intelligent optimization algorithms. Among these, the recursive least squares method suffers from underranking equations under steady-state conditions, cannot simultaneously decouple multiple electrical parameters, and requires manual pre-setting of the forgetting factor and sliding data window length. The identification performance of the signal injection method depends on the design of the injected signal amplitude and the tuning of the filter parameters, increasing the workload of design and tuning. The adaptive law of the model reference adaptive system typically adopts a proportional-integral (PI) structure, requiring proper tuning of the PI coefficients to achieve good state estimation performance. Although intelligent optimization algorithms have global search advantages, necessary parameters such as inertia weights and learning factors still require manual adjustment.

[0004] In summary, while existing state estimation methods each possess advantages in parameter identification accuracy and adaptability, their estimation performance essentially depends on the reasonable and coordinated tuning of multiple parameters, which often suffer from coupling issues. Improper parameter design can directly limit estimation accuracy and even lead to system instability, hindering practical engineering applications and widespread adoption. Summary of the Invention

[0005] To overcome the shortcomings of the prior art, this invention discloses a method for parameter estimation and robust control of permanent magnet synchronous motors.

[0006] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0007] A method for parameter estimation and robust control of a permanent magnet synchronous motor includes the following steps:

[0008] S1: Establish a finite discrete candidate set of motor parameters; based on the physical limits of the parameters to be estimated and the stability criteria of the closed-loop control system, determine the continuous global feasible region of the parameters to be estimated, and discretize the continuous global feasible region into a finite candidate set containing n elements.

[0009] S2: Construct a prediction model; Based on the discrete current prediction model of the permanent magnet synchronous motor containing parameters to be estimated, predict the current value corresponding to each candidate parameter in the finite candidate set;

[0010] S3: Design the value function; using the actual sampled current of the system as a reference, calculate the deviation between the predicted current and the actual sampled current corresponding to each candidate parameter, and construct the value function;

[0011] S4: Execute the traversal optimization algorithm; substitute each candidate parameter in the finite candidate set into the prediction model to obtain the corresponding predicted current value, then substitute the predicted current value into the value function to obtain the value function value corresponding to each candidate parameter, select the minimum value function value with the global minimum as the objective, and take the candidate parameter corresponding to the minimum value function value as the optimal estimation parameter.

[0012] S5: Substitute the optimal estimated parameters into the voltage equation of the deadbeat predictive current control to calculate the reference voltage vector at the next moment, thereby realizing closed-loop robust control of the permanent magnet synchronous motor.

[0013] Preferably, the parameter to be estimated is the stator inductance. and permanent magnet magnetic flux The continuous global feasible region is:

[0014]

[0015] The continuous global feasible region and Discretize into containing A finite candidate set of n elements and :

[0016]

[0017] in, The total number of discrete elements is determined based on the parameter estimation accuracy and computational cost.

[0018] Preferably, the permanent magnet synchronous motor uses a dq-axis discrete voltage prediction model to achieve deadbeat predictive current control, and the dq-axis discrete voltage prediction model is as follows:

[0019]

[0020] in, and The d and q axes are respectively. Predicted current at any given time; and Sample current values ​​for the d and q axes; and The stator voltages are those along the d and q axes; This is the sampling period value; This refers to the angular velocity of the motor. This is the inductance value; Stator resistance; This represents the flux linkage value of the permanent magnet.

[0021] Preferably, the prediction model constructed in S2 is: the finite candidate set The first in Candidate inductor values and the finite candidate set The first in One magnetic flux candidate value Substituting into the discrete current prediction model, we obtain the first... The d-axis predicted current corresponding to each candidate parameter and q-axis predicted current :

[0022]

[0023]

[0024] in, and They are respectively The stator current sampling value along the dq axis at time 1. and They are respectively Stator voltage at time dq axis.

[0025] Preferably, the value function in S3 and The expression is:

[0026]

[0027]

[0028] in, and For the system The actual stator sampling current at any given time is used as a reference value; and For the first The predicted current value corresponding to each candidate parameter.

[0029] Preferably, the traversal optimization algorithm in S4 specifically involves: selecting the minimum value function value along the d-axis respectively. and the minimum value function value along the q-axis Its corresponding optimal inductance and optimal flux for:

[0030]

[0031]

[0032] in, minimum value function value Corresponding candidate inductance values, minimum value function value The corresponding candidate values ​​for magnetic flux linkage.

[0033] Preferably, step S5 specifically involves: adjusting the optimal inductance... and optimal flux Substituting into the dq-axis discrete voltage prediction model, let , The reference voltage vector that needs to be applied at the next moment is calculated. and The reference voltage vector is used to generate a switching signal through a space vector pulse width modulation module to drive the power converter, thereby realizing closed-loop robust control of the permanent magnet synchronous motor.

[0034] Preferably, the traversal optimization algorithm in S4 is optimized using a binary search-based optimization algorithm, specifically by halving the continuous global feasible region proportionally. After several iterations, the continuous global feasible region converges into a narrow band interval that balances accuracy and computational cost, and the midpoint of the narrow band interval is taken as the optimal parameter output.

[0035] Preferably, the number of iterations of the bisection method The method for determining this is: based on the difference between the maximum and minimum predicted current values ​​within the parameter mismatch range. and the maximum value of current measurement error Determine the number of traversals and number of iterations The calculation formula is as follows:

[0036]

[0037]

[0038] The effective identification accuracy of the algorithm is set to be no less than the measurement accuracy of the current sensor.

[0039] Preferably, S1 to S5 are executed sequentially in each control cycle to form a parameter adaptive closed-loop control; wherein, in each control cycle, the actual stator current of the dq axis is sampled by a current sensor. and Based on the finite set value optimization, the optimal inductance and optimal flux linkage at the current moment are estimated online, and the optimal inductance and optimal flux linkage are updated in real time to the voltage equation of deadbeat predictive current control to calculate the reference voltage vector, thereby realizing robust current control under parameter mismatch conditions.

[0040] By employing the technical solution described above, the present invention has the following beneficial effects:

[0041] (1) This invention discretizes the continuous parameters to be estimated into a finite candidate set, and optimizes each candidate parameter in the finite candidate set based on the prediction model and the value function, thereby realizing the online real-time estimation of key motor parameters. The entire estimation process does not contain any adjustable parameters, and there is no need to tune the observer gain, noise covariance matrix, forgetting factor, adaptive law coefficients, etc., which fundamentally avoids the negative impact of multi-parameter coupling on estimation accuracy in existing state estimation methods, and significantly reduces the engineering implementation difficulty and debugging workload of the algorithm.

[0042] (2) This invention effectively compensates for the current prediction deviation caused by parameter mismatch by updating the optimal estimated parameters in real time to the voltage equation of the deadbeat predictive current control to calculate the reference voltage vector, and eliminates the problems of increased current ripple and steady-state error caused by parameter perturbation. Experimental results show that under the condition of double or zero-five times parameter mismatch, the proposed method can maintain good current tracking accuracy and low current harmonic distortion rate. The steady-state performance is comparable to the traditional method when the parameters are accurate, while the dynamic response speed is not affected, thus significantly improving the robustness of deadbeat predictive current control under parameter uncertainty conditions.

[0043] (3) The finite set value optimization framework proposed in this invention has good versatility and scalability. For variables to be estimated in power converters and motor drive systems that have clear boundaries and can be described by mathematical modeling, the optimization can be carried out based on this framework. It can be extended to related fields such as online optimization of controller coefficients, system bounded disturbance estimation and fault diagnosis, and has broad engineering application prospects. Attached Figure Description

[0044] Figure 1 A schematic diagram illustrating the process of establishing a finite discrete candidate set;

[0045] Figure 2 The principle of FSS-VOF and its digital implementation process (in which Figure 2 (a) shows the working principle of FSS-VOF. Figure 2 (b) is the digital implementation process);

[0046] Figure 3 The block diagram for DPCC (Dynamic Time-of-Flight Control) is shown below.

[0047] Figure 4 Here is a block diagram of DPCC based on FSS-VOF;

[0048] Figure 5 A flowchart of PMSM parameter estimation based on FSS-VOF;

[0049] Figure 6 For experimental platform;

[0050] Figure 7 The experimental results of traditional DPCC under the conditions of 500 r / min and 5 N·m are accurate.

[0051] Figure 8 Experimental results of the two methods under the conditions of 500 r / min, 5 N·m and a parameter mismatch of 2 times;

[0052] Figure 9 Experimental results of the two methods under the conditions of 500 r / min, 5 N·m and 0.5 times parameter mismatch;

[0053] Figure 10 The experimental results are presented when switching between the two methods at 1000 r / min and 5 N·m (2 times parameter mismatch).

[0054] Figure 11 The experimental results are shown when switching between the two methods at 1000 r / min and 5 N·m (0.5 times parameter mismatch).

[0055] Figure 12 Comparison of phase current THD values ​​for two methods under 2x parameter mismatch, rated load torque, and different speeds;

[0056] Figure 13 A diagram showing the dynamic results of a traditional DPCC when the parameters are accurate;

[0057] Figure 14 A dynamic result graph when the parameters are accurate;

[0058] Figure 15 The dynamic results are shown under a 2x parameter mismatch.

[0059] Figure 16 The dynamic results are shown under a parameter mismatch of 0.5. Detailed Implementation

[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0061] Combined with appendix Figures 1-16 The present invention discloses a method for parameter estimation and robust control of a permanent magnet synchronous motor in some embodiments, comprising the following steps:

[0062] S1: Establish a finite discrete candidate set of motor parameters; based on the physical limits of the parameters to be estimated and the stability criteria of the closed-loop control system, determine the continuous global feasible region of the parameters to be estimated, and discretize the continuous global feasible region into a set containing... A finite set of candidates for each element;

[0063] S2: Construct a prediction model; Based on the discrete current prediction model of the permanent magnet synchronous motor containing parameters to be estimated, predict the current value corresponding to each candidate parameter in the finite candidate set;

[0064] S3: Design the value function; using the actual sampled current of the system as a reference, calculate the deviation between the predicted current and the actual sampled current corresponding to each candidate parameter, and construct the value function;

[0065] S4: Execute the traversal optimization algorithm; substitute each candidate parameter in the finite candidate set into the prediction model to obtain the corresponding predicted current value, then substitute the predicted current value into the value function to obtain the value function value corresponding to each candidate parameter, select the minimum value function value with the global minimum as the objective, and take the candidate parameter corresponding to the minimum value function value as the optimal estimation parameter.

[0066] S5: Substitute the optimal estimated parameters into the voltage equation of the deadbeat predictive current control to calculate the reference voltage vector at the next moment, thereby realizing closed-loop robust control of the permanent magnet synchronous motor.

[0067] Specifically, the core idea of ​​FCS-MPC is to utilize a finite set of voltage vectors from the converter, predict the trajectory of each candidate voltage vector in the future using a system model, and select the optimal voltage vector to apply to the next control cycle based on a value function. FSS-VOF discretizes continuous parameters or states to be estimated into a finite set, calculates the predicted value corresponding to each candidate state in the set using a system model, and selects the optimal state from the finite set based on a constructed value function. The core architecture of FSS-VOF consists of four parts: a finite discrete candidate set, a prediction model, a value function, and an traversal optimization algorithm. (Refer to...) Figure 2 , Figure 2 (a) shows the working principle of FSS-VOF, in which the controller and the main circuit constitute the controlled object. The FSS-VOF module sequentially performs four steps: finite discrete candidate set construction, prediction model calculation, value function evaluation and traversal optimization. Finally, it outputs the optimal estimated state and feeds it back to the controller to realize the closed-loop control of the system. Figure 2 (b) demonstrates the digital implementation process of FSS-VOF, which involves discretizing the continuous feasible domain into a finite candidate set, and then performing digital operations on the prediction model, value function, and ergonomic optimization to output the optimal estimated state.

[0068] Reference Figure 3 This is a control block diagram for deadbeat predictive current control (DPCC). In this system, the outer speed loop uses a PI controller, which calculates the difference between the given speed and the actual motor speed, and outputs the q-axis reference current after calculation. d-axis reference current Set to 0. The inner current loop is a deadbeat predictive current control module, based on the dq-axis reference current and the sampled actual dq-axis current. , The reference voltage is calculated by combining the mathematical model of the motor. , Then, the switching signal is generated through coordinate transformation and space vector pulse width modulation (SVPWM) modules. , , The power converter drives the permanent magnet synchronous motor (PMSM) to operate. Current sensors sample the three-phase current. The dq-axis current is obtained through coordinate transformation. , As feedback.

[0069] The permanent magnet synchronous motor DPCC utilizes the dq-axis discrete voltage prediction model shown in equation (6) to predict the voltage of the dq axis. Current prediction at time , Closely follow the given reference value , (that is, to believe) , This enables fast current command tracking.

[0070]

[0071] in: and The d and q axes are respectively. Predicted current at any given time; and Sample current values ​​for the d and q axes; and The stator voltages are those along the d and q axes; This is the sampling period value; This refers to the angular velocity of the motor. This is the inductance value; Stator resistance; This represents the flux linkage value of the permanent magnet.

[0072] According to equation (6), the expression for the discrete current prediction equation under the nominal motor parameters is as follows:

[0073]

[0074] In the formula, , as well as These are the nominal values ​​for the motor parameters.

[0075] However, in actual operation of a motor, parameter mismatch often occurs. The relationship between the motor's nominal values ​​and actual parameters can be expressed as:

[0076]

[0077] in: , as well as Represents actual motor parameters; , as well as This represents the perturbation value of the motor parameters, and its magnitude is the difference between the actual value and the nominal value of the motor parameters.

[0078] Based on the current prediction model expressed by equation (7), if the parameter mismatch problem in the model is considered, the real current prediction model can be expressed as:

[0079]

[0080] Subtracting equations (7) and (9) yields the expression for the current error caused by parameter mismatch:

[0081]

[0082] As shown in equation (10), the d-axis stator current is affected by the mismatch between resistance and inductance parameters, while the q-axis stator current is affected by the mismatch between resistance, inductance, and flux linkage. This will cause a steady-state error in the current loop. Among them, flux linkage mismatch only affects the q-axis current; inductance mismatch affects both the d-axis and q-axis currents. It is worth noting that resistance mismatch usually has a small impact and can be ignored.

[0083] Reference Figure 4 The diagram below shows the block diagram of a DPCC based on FSS-VOF. In this parameter-robust control method, FSS-VOF is used to estimate motor parameters in real time. Its design strictly follows four steps: "finite discrete candidate set, prediction model, value function, and ergonomic optimization algorithm design". Compared with traditional DPCC, this scheme adds a parameter estimation module based on FSS-VOF, which receives the sampled current of the dq axis. , and voltage As input, the optimal estimated parameters are output to the deadbeat predictive current control module, thereby achieving robust control with parameter adaptation.

[0084] In some embodiments, the parameter to be estimated is the stator inductance. and permanent magnet magnetic flux The continuous global feasible region is:

[0085]

[0086] The continuous global feasible region and Discretize into containing A finite candidate set of n elements and :

[0087]

[0088] in, The total number of discrete elements is determined based on the parameter estimation accuracy and computational cost.

[0089] Specifically, given the motor resistance The perturbation has a relatively small impact on system performance, therefore only the stator inductor needs to be adjusted. and permanent magnet magnetic flux Real-time parameter estimation is performed. To ensure system stability, the inductance parameter mismatch range is 0 to 2 times the actual inductance; therefore, the upper and lower limits of the inductance to be estimated are set as follows: Similarly, the lower limit of the flux linkage is set to 0, and the upper limit is set according to the standard of 200% parameter mismatch. Therefore, the continuous global feasible region of the parameters to be estimated is shown in equation (11). The above continuous global feasible region... and Discretize into containing A finite candidate set of n elements and As shown in equation (12). In the equation, the total number of discrete elements... This determines the accuracy of the motor parameter estimation. A larger value results in higher precision, but also leads to more subsequent traversals and increased computational complexity. Therefore, the value should be determined based on system requirements. (See reference...) Figure 1 This diagram illustrates the process of establishing a finite discrete candidate set, showing how the continuous feasible region is constructed from... arrive Uniform discretization element The process involves each discrete element corresponding to a candidate parameter value in the candidate set.

[0090] In some embodiments, the permanent magnet synchronous motor employs a dq-axis discrete voltage prediction model to achieve deadbeat predictive current control, and the dq-axis discrete voltage prediction model is shown in equation (6). This model is the basis for deadbeat predictive current control and also the model foundation for constructing the prediction model in subsequent parameter estimation.

[0091] In some embodiments, the prediction model constructed in S2 is: the finite candidate set The first in Candidate inductor values and the finite candidate set The first in One magnetic flux candidate value Substituting into the discrete current prediction model, we obtain the first... The d-axis predicted current corresponding to each candidate parameter and q-axis predicted current :

[0092]

[0093] In the formula, Representing a finite set The first in One inductor candidate value; Representing a finite set The first in One magnetic flux candidate value, and They are respectively The stator current sample value at time dq axis. As can be seen from formula (13), when the inductance and flux linkage take different candidate values, the corresponding predicted current value will also be different. This prediction model is constructed based on the current prediction model shown in formula (7). Since the current prediction model includes the inductance and flux linkage parameters to be estimated, the discrete candidate set can be predicted based on this model. and The current values ​​corresponding to each inductor and magnetic flux are determined, thus laying the model foundation for subsequent value function evaluation.

[0094] In some embodiments, the value function in S3 and The expression is:

[0095]

[0096] in, and For the system The actual stator sampling current at any given time is used as a reference value; and For the first The predicted current value corresponding to each candidate parameter.

[0097] From formula (13), it can be seen that under voltage and Under the influence of [the system / mechanical parameters], the actual current sampled value of the motor reflects the true motor parameter information. This means that when the inductance and flux linkage are true parameters, the corresponding predicted current value should be the same as the actual current sampled value. If the inductance and flux linkage values ​​are not true parameters, the corresponding predicted current value will deviate from the actual current sampled value. [This is to] select a finite candidate set. and The optimal parameters need to be determined by quantifying how closely each candidate parameter approximates the true parameters. This embodiment sets the system... The actual stator sampling current at any given moment and As a reference benchmark, a value function is designed to evaluate the predicted current corresponding to each candidate parameter. When the candidate parameter in the finite set differs from the actual motor parameter, the predicted current value of that candidate parameter will deviate from the sampled current value (this deviation is determined by the value function value). and (This is expressed as follows). The closer the candidate parameters are to the actual parameters of the motor, the smaller the deviation between the predicted current value and the sampled current value. and (The smaller the value). Therefore, by determining the magnitude of the deviation between the predicted current value and the sampled current value of each candidate parameter, the optimal candidate parameter can be determined.

[0098] In some embodiments, the traversal optimization algorithm in S4 specifically involves: selecting the minimum value function value along the d-axis respectively. and the minimum value function value along the q-axis Its corresponding optimal inductance and optimal flux for:

[0099]

[0100] in, minimum value function value Corresponding candidate inductance values, minimum value function value The corresponding candidate values ​​for magnetic flux linkage.

[0101] finite candidate set and Each candidate parameter is substituted into the constructed current prediction model (13) to obtain the current prediction value corresponding to each candidate parameter, and then substituted into the value function. and In the middle, we obtained Each has a value function value. With the goal of finding the "global minimum," the minimum value function value is selected from each. and The discrete inductance and flux linkage candidate values ​​corresponding to this minimum value function are the optimal inductance. and optimal flux . Reference Figure 5 The diagram shows the parameter estimation process for a PMSM based on FSS-VOF. It illustrates the complete parameter estimation procedure: First, a finite discrete candidate set is established. Then, by iterating through the loop "For i=1 ton", each candidate parameter is sequentially substituted into the predictive model to calculate the predicted current. Finally, the cost function is calculated. Then, the traversal optimization phase begins, selecting the... , Output the optimal estimated state , .

[0102] In some embodiments, S5 specifically involves: adjusting the optimal inductance... and optimal flux Substituting into the dq-axis discrete voltage prediction model, let , The reference voltage vector that needs to be applied at the next moment is calculated. and The reference voltage vector is used to generate a switching signal through a space vector pulse width modulation module to drive the power converter, thereby realizing closed-loop robust control of the permanent magnet synchronous motor.

[0103] Finally, the estimated optimal inductance and flux linkage values ​​are substituted into the voltage equation (6) to calculate the reference voltage vector to be applied at the next moment. This vector is then sent to the SVPWM module to generate a switching signal for the power converter, thereby achieving closed-loop control. Since the parameters used in the control law are optimal values ​​obtained through online estimation using finite set value optimization, they can effectively compensate for the current tracking error caused by parameter mismatch and achieve robust current control.

[0104] In some embodiments, the traversal optimization algorithm in S4 is optimized using a bisection-based optimization algorithm, specifically by halving the continuous global feasible region proportionally. After several iterations, the continuous global feasible region converges into a narrow band interval that balances accuracy and computational cost, and the midpoint of the narrow band interval is taken as the optimal parameter output.

[0105] Although the traversal optimization algorithm described above enumerates all The method can obtain the global optimal solution by selecting several candidate parameters, but it requires execution in each control cycle. The sub-value function calculation and comparison operations, when high control frequency or resolution requirements are needed, involve increasing the number of candidate set elements. The increase in computation will directly consume the remaining computing power within the interrupt service routine, affecting the execution of other real-time tasks. To reduce computation while ensuring the accuracy of global optimization, this embodiment introduces a bisection optimization strategy to optimize the traversal optimization algorithm. Its core idea is to define the continuous global feasible region of motor parameters. and The process involves "proportionally halving" the value. After... In the next iteration, the feasible region and The algorithm converges to a narrow band interval that balances accuracy and computational complexity, and finally takes the midpoint of the interval as the optimal parameter output.

[0106] In some embodiments, the number of iterations of the bisection method The method for determining this is: based on the difference between the maximum and minimum predicted current values ​​within the parameter mismatch range. and the maximum value of current measurement error Determine the number of traversals and number of iterations The calculation formula is as follows:

[0107]

[0108]

[0109] The effective identification accuracy of the algorithm is set to be no less than the measurement accuracy of the current sensor.

[0110] Since the core value function of FSS-VOF is constructed based on the prediction error of the stator current, the effective identification accuracy of the algorithm is physically determined by the hardware error of the current sampling. That is, the algorithm's identification accuracy should not be lower than the measurement accuracy of the current sensor. This embodiment uses an LA-50P model current Hall sensor with a turns ratio of 2 and a minimum measurement accuracy of 0.5%. Therefore, the maximum current measurement error can be determined to be... A. At this level of precision, the number of traversals required by the traversal optimization algorithm is: However, using the binary search method only requires 6 optimization attempts. The binary search algorithm can greatly reduce the amount of computation while ensuring the accuracy of the search.

[0111] Specific number of traversals and number of iterations The calculation process is as follows. The maximum value of the inductance under parameter mismatch conditions can be determined using formula (11). and minimum value The values ​​are 0.0132mH and 0.001mH respectively (to ensure the denominator is not zero, the minimum inductance is set to 0.001mH); the maximum value of the magnetic flux. and minimum value The values ​​are 0.48 Wb and 0 Wb, respectively. Meanwhile, the dq-axis voltage and current under rated conditions can be determined by voltage equation (6), which yields... A, A (Inductance flux parameters are based on motor nameplate data).

[0112] Based on the above data, the maximum predicted current value, minimum predicted current value, and the difference between the maximum and minimum predicted current values ​​of the dq axis are calculated for three mismatch conditions: "inductance mismatch", "magnetic flux mismatch", and "both inductance and magnetic flux mismatch".

[0113] When inductance parameters are mismatched, according to formula (7), the maximum and minimum predicted current values ​​along the d-axis are respectively:

[0114]

[0115]

[0116] The maximum and minimum predicted current values ​​along the q-axis are as follows:

[0117]

[0118]

[0119] Therefore, the difference between the maximum and minimum predicted currents along the dq axis can be obtained. , They are respectively:

[0120]

[0121]

[0122] When flux linkage parameters are mismatched, the prediction model for the d-axis stator current does not contain a flux linkage term. Flux linkage mismatch does not affect the d-axis stator current; therefore, the difference between the maximum and minimum values ​​of the predicted d-axis current is not significant. A. The maximum and minimum predicted current values ​​along the q-axis are respectively:

[0123]

[0124]

[0125] Therefore, the difference between the maximum and minimum values ​​of the predicted q-axis current. for:

[0126]

[0127] When both flux linkage and inductance are mismatched, the d-axis stator current is only related to inductance mismatch. Therefore, the difference between the maximum and minimum predicted d-axis currents when both inductance and flux linkage are mismatched is the same as that when inductance is mismatched. A. The maximum and minimum predicted current values ​​along the q-axis are respectively:

[0128]

[0129]

[0130] Therefore, the difference between the maximum and minimum values ​​of the predicted q-axis current. for:

[0131]

[0132] By comparing the difference between the maximum and minimum predicted currents of the dq axis in the three cases, i.e., selecting the maximum value from formulas (20), (21), (24), and (27), it can be seen that... A.

[0133] Number of traversals With the number of iterations The selections are all related to and the maximum value of current measurement error The relevant calculation formulas are shown in equations (28) and (29). From equations (28) and (29), it can be seen that, under the same precision, the number of traversals required when using the traversal method is [missing information]. The number of iterations required by the bisection method .

[0134] In some embodiments, the method sequentially executes S1 to S5 in each control cycle to form a parameter adaptive closed-loop control; wherein, in each control cycle, the actual stator current of the dq axis is sampled by a current sensor. and Based on the finite set value optimization, the optimal inductance and optimal flux linkage at the current moment are estimated online, and the optimal inductance and optimal flux linkage are updated in real time to the voltage equation of deadbeat predictive current control to calculate the reference voltage vector. Robust current control under parameter mismatch conditions can be achieved without additional PI controllers, observers or adaptive laws.

[0135] The entire architecture does not contain any adjustable parameters during implementation; therefore, the proposed FSS-VOF has the advantage of being parameter-free. Compared with existing state estimation methods (such as sliding mode observers requiring tuning of observer and reaching law parameters, extended Kalman filters requiring tuning of process noise and measurement noise covariance matrices, recursive least squares requiring tuning of forgetting factors and initial covariance values, and model reference adaptation requiring tuning of PI coefficients), FSS-VOF requires no parameter tuning, avoiding complex parameter tuning processes and solving the problem of multi-parameter coupling affecting estimation accuracy in existing methods. It is important to emphasize that FSS-VOF is not limited to the estimation of parameters or state variables; for variables with boundaries in power converters and motor drive systems that can be described by mathematical modeling, FSS-VOF can be used for optimal selection.

[0136] To verify the effectiveness of the proposed method, in the following... Figure 6 Steady-state and dynamic experiments under different operating conditions were conducted on the experimental platform shown. The experimental platform includes an AC voltage regulator, oscilloscope, isolation transformer, load motor, drive motor, host computer, ControlDesk, MicroLabBox DS1202, drive circuit, DC power supply, current and speed sampling modules, series load port, and DSP controller. The parameters of the surface-mount permanent magnet synchronous motor (SPMSM) used in the experiment are: DC bus voltage... V, rated power kW, stator resistance Rated torque N·m, inductance mH, rated speed r / min, permanent magnet flux Wb, polar logarithm Moment of inertia kg·m², sampling period s.

[0137] Reference Figure 7 The results are experimental results of a conventional DPCC under conditions of 500 r / min and 5 N·m (parameters accurate). It can be seen that the ripple of the d-axis and q-axis currents are 0.4 A and 0.6 A, respectively, and the total harmonic distortion (THD) of the phase current is 3.57%. Furthermore, the q-axis sampling current can effectively track the given q-axis value.

[0138] Reference Figure 8 The results are experimental results of the two methods under the conditions of 500 r / min, 5 N·m and a parameter mismatch of 2 times. Figure 8 (a) demonstrates the limitations of the traditional DPCC method in addressing a 2x mismatch in motor parameters ( , The experimental results at the specified time show that both the phase current and the dq-axis stator current exhibit significant distortion, with a total harmonic distortion (THD) of 26.09%. Furthermore, the dq-axis current ripple is 2.4A and 3.4A respectively, which are larger than the ripple values ​​when the parameters are accurate. This indicates that the traditional method leads to a severe deterioration in the steady-state performance of the PMSM system when the parameters are mismatched by a factor of two. In contrast, Figure 8 (b) shows the experimental results of the proposed method under the condition of 2 times parameter mismatch. It can be seen that the THD of the proposed method is 3.12%, which is much smaller than that of the traditional DPCC at 26.09%. The current ripple of the dq axis is also reduced to 0.6A and 0.4A. Its steady-state performance is similar to that of the traditional DPCC under the condition of accurate parameters. This shows that the proposed method can effectively solve the problem of large current ripple caused by parameter mismatch.

[0139] Reference Figure 9 The results are experimental results of the two methods under the conditions of 500 r / min, 5 N·m and 0.5 times parameter mismatch. Figure 9 (a) demonstrates the performance of a conventional DPCC with a motor parameter mismatch of 0.5 times ( , The experimental results show that there is a significant steady-state error between the q-axis current sampling value and the q-axis current setpoint of the traditional DPCC. Figure 9 (b) shows the experimental results of the proposed method under 0.5 times parameter mismatch. It can be seen that there is no steady-state error between the q-axis sampling current and the q-axis current setpoint, which indicates that the proposed method still has good current tracking capability under 0.5 times parameter mismatch.

[0140] Reference Figure 10The figures show the experimental results (2x parameter mismatch) when switching between the two methods at 1000 r / min and 5 N·m. It can be seen that after switching to the proposed method, the total harmonic distortion (THD) of the phase current decreases from 18.56% to 3.21%. (Refer to...) Figure 11 The results are experimental findings (0.5x parameter mismatch) when switching between the two methods at 1000 r / min and 5 N·m. Figure 11 The results show that, under a parameter mismatch of 0.5 times, the significant current static error is eliminated after switching from the traditional DPCC to the proposed method, further verifying the superiority of the proposed method.

[0141] Reference Figure 12 The comparison shows the phase current THD values ​​of the two methods under conditions of 2x parameter mismatch, rated load torque, and different speeds. It can be seen that under different speeds and 2x parameter mismatch, the phase current of the proposed method can still maintain a good THD value.

[0142] The steady-state experimental results above show that the proposed method can maintain good steady-state performance under different parameter mismatch conditions, thus avoiding the negative impact of parameter mismatch on the system control performance.

[0143] Reference Figure 13 The figure shows the dynamic results of a traditional DPCC when the parameters are accurate. The experimental conditions were a rotational speed of 1000 r / min, a control frequency of 16 kHz, and a sudden change in load torque from 3 N·m to 5 N·m. (Refer to...) Figure 14 The figure shows the dynamic results of the proposed method when the parameters are accurate. The experimental results show that the dynamic response time of the proposed method is almost the same as that of the traditional method without parameter mismatch, which means that the proposed method also has satisfactory dynamic performance.

[0144] Reference Figure 15 The figure shows the dynamic results of the proposed method under a 2x parameter mismatch. (Refer to...) Figure 16 The figure shows the dynamic results of the proposed method under 0.5 times parameter mismatch. It can be seen that the proposed method also has a fast dynamic response, with the response time being the same as that without parameter mismatch, proving that the proposed method still has good dynamic performance under parameter mismatch conditions.

[0145] In summary, the experimental results validate the effectiveness of the proposed FSS-VOF. Furthermore, the DPCC method based on FSS-VOF achieves good steady-state and dynamic control performance under different parameter mismatch conditions and operating states, solving the problem of poor parameter robustness in DPCC. The proposed FSS-VOF can estimate state variables online in real time without requiring any adjustment parameters, making it easy to implement. Therefore, this method has significant value for engineering applications and technology promotion. Based on its general finite state set value optimization architecture, it can be extended to more research areas, including online optimization of controller coefficients in power converters and motor drive systems, bounded disturbance estimation of systems, and fault diagnosis of power converters and motor drive systems.

[0146] It should be noted that the components or steps in the above embodiments can be interchanged, substituted, added, or deleted. Therefore, the combinations formed by these reasonable permutations and transformations should also fall within the protection scope of this invention, and the protection scope of this invention should not be limited to the above embodiments.

[0147] The parts of this invention not described in detail are prior art. It will be apparent to those skilled in the art that this invention is not limited to the details of the above exemplary embodiments, and that the invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects, and are intended to encompass all changes falling within the meaning and scope of equivalents within this invention.

Claims

1. A method for parameter estimation and robust control of a permanent magnet synchronous motor, characterized in that, Includes the following steps: S1: Establish a finite discrete candidate set of motor parameters; based on the physical limits of the parameters to be estimated and the stability criteria of the closed-loop control system, determine the continuous global feasible region of the parameters to be estimated, and discretize the continuous global feasible region into a finite candidate set containing n elements. S2: Construct a prediction model; Based on the discrete current prediction model of permanent magnet synchronous motor containing parameters to be estimated, the current value corresponding to each candidate parameter in the finite candidate set is predicted. S3: Design the value function; Using the actual sampled current of the system as a reference, the deviation between the predicted current and the actual sampled current corresponding to each candidate parameter is calculated, and a value function is constructed. S4: Execute the traversal optimization algorithm; The candidate parameters in the finite candidate set are substituted into the prediction model in turn to obtain the corresponding predicted current value. Then, the predicted current value is substituted into the value function to obtain the value function value corresponding to each candidate parameter. The minimum value function value is selected with the goal of global minimization. The candidate parameter corresponding to the minimum value function value is used as the optimal estimation parameter. S5: Substitute the optimal estimated parameters into the voltage equation of the deadbeat predictive current control to calculate the reference voltage vector at the next moment, thereby realizing closed-loop robust control of the permanent magnet synchronous motor.

2. The parameter estimation and robust control method for permanent magnet synchronous motors as described in claim 1, characterized in that, The parameter to be estimated is the stator inductance. and permanent magnet magnetic flux The continuous global feasible region is: The continuous global feasible region and Discretize into containing A finite candidate set of n elements and : in, The total number of discrete elements is determined based on the parameter estimation accuracy and computational cost.

3. The parameter estimation and robust control method for permanent magnet synchronous motors as described in claim 2, characterized in that, The permanent magnet synchronous motor uses a dq-axis discrete voltage prediction model to achieve deadbeat predictive current control. The dq-axis discrete voltage prediction model is as follows: in, and The d and q axes are respectively. Predicted current at any given time; and Sample current values ​​for the d and q axes; and The stator voltages are those along the d and q axes; This is the sampling period value; This refers to the angular velocity of the motor. This is the inductance value; Stator resistance; This represents the flux linkage value of the permanent magnet.

4. The parameter estimation and robust control method for permanent magnet synchronous motors as described in claim 3, characterized in that, The prediction model constructed in S2 is: the finite candidate set The first in Candidate inductor values and the finite candidate set The first in One magnet linkage candidate value Substituting into the discrete current prediction model, we obtain the first... The d-axis predicted current corresponding to each candidate parameter and q-axis predicted current : in, and They are respectively The stator current sampling value along the dq axis at time 1. and They are respectively Stator voltage at time dq axis.

5. The parameter estimation and robust control method for permanent magnet synchronous motors as described in claim 4, characterized in that, The value function in S3 and The expression is: in, and For the system The actual stator sampling current at any given time is used as a reference value; and For the first The predicted current value corresponding to each candidate parameter.

6. The parameter estimation and robust control method for permanent magnet synchronous motors as described in claim 5, characterized in that, The traversal optimization algorithm in S4 specifically involves: selecting the minimum value function value along the d-axis. and the minimum value function value along the q-axis Its corresponding optimal inductance and optimal flux for: in, minimum value function value Corresponding candidate inductance values, minimum value function value The corresponding candidate values ​​for magnetic flux linkage.

7. The parameter estimation and robust control method for permanent magnet synchronous motors as described in claim 6, characterized in that, Specifically, S5 involves: adjusting the optimal inductance... and optimal flux Substituting into the dq-axis discrete voltage prediction model, let , The reference voltage vector that needs to be applied at the next moment is calculated. and The reference voltage vector is used to generate a switching signal through a space vector pulse width modulation module to drive the power converter, thereby realizing closed-loop robust control of the permanent magnet synchronous motor.

8. The parameter estimation and robust control method for permanent magnet synchronous motors as described in claim 1, characterized in that, The traversal optimization algorithm in S4 is optimized using a bisection-based optimization algorithm, specifically by halving the continuous global feasible region proportionally. After one iteration, the continuous global feasible region converges into a narrow band interval that balances accuracy and computational cost, and the midpoint of the narrow band interval is taken as the optimal parameter output.

9. The parameter estimation and robust control method for permanent magnet synchronous motors as described in claim 8, characterized in that, The number of iterations of the bisection method The method for determining this is: based on the difference between the maximum and minimum predicted current values ​​within the parameter mismatch range. and the maximum value of current measurement error Determine the number of traversals and number of iterations The calculation formula is as follows: The effective identification accuracy of the algorithm is set to be no less than the measurement accuracy of the current sensor.

10. The parameter estimation and robust control method for permanent magnet synchronous motors as described in claim 1, characterized in that, S1 to S5 are executed sequentially in each control cycle to form a parameter adaptive closed-loop control; in each control cycle, the actual stator current of the dq axis is sampled by a current sensor. and Based on the finite set value optimization, the optimal inductance and optimal flux linkage at the current moment are estimated online, and the optimal inductance and optimal flux linkage are updated in real time to the voltage equation of deadbeat predictive current control to calculate the reference voltage vector, thereby realizing robust current control under parameter mismatch conditions.

Images