A speed control method and system for permanent magnet synchronous motor model prediction

Abstract

The present invention discloses a speed control method and system for a permanent magnet synchronous motor based on model prediction. The method includes steps such as parameter measurement, model correction, model prediction, cost function minimization, output drive, etc. The optimal voltage vector is output to the inverter to drive the motor to operate, realizing the direct control of the motor speed and eliminating the influence of parameter uncertainty on the control effect. The method provided by the embodiments of the present application corrects the accuracy of the model based on a load observer and a moment of inertia identification algorithm by considering the prediction error problem caused by parameter uncertainty. At the same time, by introducing a motor PI cost term into the cost function, the steady-state error under the condition of parameter mismatches such as inductance and permanent magnet flux linkage is eliminated, and finally the parameter robustness control is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of motor control, and particularly to a speed control method and system for model prediction of a permanent magnet synchronous motor. Background Art

[0002] With the wide application of servo drives in industry, the requirements for the dynamic performance and reliability of servo drive speed control are getting higher and higher. In the traditional speed control of a permanent magnet synchronous motor (PMSM), the most commonly used is double closed-loop cascade control, with the outer loop being the speed loop and the inner loop being the current loop. For traditional cascade linear controllers such as vector control, direct torque control, and finite set model predictive control, in order to avoid excessive overshoot, it is necessary to limit the bandwidths of their inner and outer loops to make them match each other. Also, due to its cascade structure, it brings problems such as greater difficulty in tuning the proportional-integral parameters and low dynamic response.

[0003] To solve the above problems, researchers in the prior art have proposed to improve the control performance by means of new control structures and control methods. For example, Ming Liu et al. (Ming Liu, Ka Wing Chan, Jiefeng Hu et al. Model Predictive Direct Speed Control With Torque Oscillation Reduction for PMSM Drives [J]. IEEE Transactions on Industrial Informatics, 2019, 15(9): 4944 - 4956) proposed a model predictive direct speed control (MPDSC), which adopts a single-loop structure, predicts the future speed through discrete equations, eliminates the cascade structure, and then selects the optimal voltage vector for controlling the motor according to a cost function based on speed and flux linkage, etc., which is beneficial to fully exert the dynamic response ability of the PMSM permanent magnet synchronous motor.

[0004] However, despite the significant progress made in the research of MPDSC, the challenge of strong dependence on the system model remains to be solved. In model predictive control, the mathematical model of the system is used to predict the values of state variables. There are problems of parameter mismatch and model uncertainty in the control performance of model predictive control, such as the jump of load torque, the parameter uncertainty of moment of inertia, the fluctuation of inductance and permanent magnet flux linkage parameters, etc. The motor parameters in the model of MPDSC are crucial for the prediction performance, which means that the mismatch of all these parameters will lead to prediction errors in the control behavior of the machine, resulting in the control performance not reaching the predicted effect, such as the existence of steady-state error in speed and the instability of the system. As a control method based on the motor model, it is very necessary to improve the parameter robustness of model predictive direct speed control.

[0005] Therefore, how to solve the problem of parameter robustness of model predictive direct speed control is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] The present invention provides a speed control method and system for model prediction of a permanent magnet synchronous motor.

[0007] The present invention provides the following solutions:

[0008] A speed control method for model prediction of a permanent magnet synchronous motor, comprising:

[0009] Measuring the parameters of the permanent magnet synchronous motor at a certain moment through a sensor, where the parameters at least include voltage, current and speed; Obtaining an estimated value of the load torque in the prediction model through a reduced-order Luenberger load observer;

[0010] Obtaining an estimated value of the moment of inertia in the prediction model through a model reference moment of inertia identification algorithm;

[0011] Feeding back the estimated value of the load torque and the estimated value of the uncertain moment of inertia to the prediction model to realize the correction of model uncertainty, and using the estimated value of the load torque as the reference value of the electromagnetic torque;

[0012] Predicting the current values at time +1 and +2 through the corrected motor discrete model prediction equation and one-step delay compensation, and further obtaining the predicted values of speed, magnetic flux and electromagnetic torque;

[0013] Inputting the predicted values of the speed, magnetic flux and electromagnetic torque, as well as the given speed, magnetic flux and electromagnetic torque, into the cost function; together with the finite set of voltage vectors, an optimal voltage vector is selected to determine the optimal voltage vector, and the optimal voltage vector minimizes the cost function; +1, +2

[0014] Inputting the predicted values of the speed, magnetic flux, and electromagnetic torque, as well as the given speed, magnetic flux, and electromagnetic torque, into the cost function; together with the finite set of voltage vectors, an optimal voltage vector is selected to determine the optimal voltage vector, and the optimal voltage vector minimizes the cost function;

[0015] Output the optimal voltage vector to the inverter to drive the motor to operate.

[0016] A speed control system for a permanent magnet synchronous motor based on model prediction, the system comprising:

[0017] A reference acquisition unit for measuring the parameters of the permanent magnet synchronous motor at a moment, the parameters at least including voltage, current and rotational speed;

[0018] A load torque estimation unit for obtaining an estimated value of the load torque in the prediction model through a reduced-order Luenberger load observer;

[0019] A moment of inertia estimation unit for obtaining an estimated value of the moment of inertia in the prediction model through a model reference moment of inertia identification algorithm;

[0020] A model correction unit for feeding back the load torque estimated value and the uncertain moment of inertia estimated value to the prediction model, realizing the correction of model uncertainty, and using the load torque estimated value as a reference value of the electromagnetic torque;

[0021] A compensation unit for obtaining +1, the current values at +2 moments through the corrected motor discrete model prediction equation and one-step delay compensation, and further obtaining the predicted values of rotational speed, magnetic flux and electromagnetic torque;

[0022] An optimal voltage vector selection unit for inputting the predicted values of the rotational speed, magnetic flux and electromagnetic torque and the given rotational speed, magnetic flux and electromagnetic torque into a cost function; together with a finite set of voltage vectors, performing optimal voltage vector selection to determine the optimal voltage vector, the optimal voltage vector making the cost function minimum;

[0023] An optimal voltage vector output unit for outputting the optimal voltage vector to the inverter to drive the motor to operate.

[0024] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0025] The method provided by the embodiments of the present application corrects the accuracy of the model by considering the prediction error problem caused by parameter uncertainty, based on a load observer and a moment of inertia identification algorithm. At the same time, by introducing a motor PI cost term into the cost function, the elimination of the steady-state error in the case of parameter mismatches such as inductance and permanent magnet chain is realized, and finally the parameter robustness control is achieved.

[0026] Of course, any product implementing the present invention does not necessarily need to achieve all the above-mentioned advantages simultaneously. Description of the Drawings

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0028] Figure 1 is a flowchart of a speed control method for model prediction of a permanent magnet synchronous motor provided by an embodiment of the present invention;

[0029] Figure 2 is a structural block diagram of model predictive direct speed control provided by an embodiment of the present invention;

[0030] Figure 3 is a schematic diagram of one-step delay compensation prediction provided by an embodiment of the present invention;

[0031] Figure 4 is a flowchart of model predictive direct speed control provided by an embodiment of the present invention;

[0032] Figure 5 is a schematic diagram of parameter sensitivity provided by an embodiment of the present invention;

[0033] Figure 6 is another schematic diagram of parameter sensitivity provided by an embodiment of the present invention;

[0034] Figure 7 is an observation effect diagram of a load torque observer provided by an embodiment of the present invention;

[0035] Figure 8 is an observation effect diagram of a moment of inertia identification algorithm provided by an embodiment of the present invention;

[0036] Figure 9(a) is an effect diagram of control for changes in inductance parameters (original MPDSC) provided by an embodiment of the present invention;

[0037] Figure 9(b) is an effect diagram of control for changes in inductance parameters (improved MPDSC) provided by an embodiment of the present invention;

[0038] Figure 10(a) is an effect diagram of control for changes in permanent magnet chain parameters (original MPDSC) provided by an embodiment of the present invention;

[0039] Figure 10(b) is an effect diagram of control for changes in permanent magnet chain parameters (improved MPDSC) provided by an embodiment of the present invention;

[0040] Figure 11(a) is an effect diagram of control for changes in moment of inertia parameters (original MPDSC) provided by an embodiment of the present invention;

[0041] Figure 11(b) is the control effect diagram of the change of the moment of inertia parameter provided by the embodiment of the present invention (improved MPDSC);

[0042] Figure 12 is a schematic diagram of a speed control system for a permanent magnet synchronous motor model prediction provided by the embodiment of the present invention. Specific embodiments

[0043] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art belong to the scope of protection of the present invention.

[0044] This method is based on the control method of model predictive direct speed control, abandons the cascade structure of double closed-loop control, avoids the complex parameter tuning process, reduces overshoot, and improves the speed response. Analyze the parameter sensitivity of the MPDSC method, point out that parameter mismatch will lead to prediction speed error and inaccurate delay compensation, and solve the problems brought by parameter uncertainty through a load torque observer, a moment of inertia identification algorithm, and a proportional integral (PI) cost function, improving the robustness of model predictive direct speed control.

[0045] This method considers the necessary motor parameter uncertainties. Based on model predictive direct speed control, while improving the response speed, through a load torque observer and a moment of inertia identification algorithm, the accuracy of the prediction model is improved. By introducing the speed PI cost function into the original cost function, the steady-state error caused by parameter mismatch is eliminated, and the robustness of model predictive direct speed control is improved.

[0046] See Figure 1 , a speed control method for a permanent magnet synchronous motor model prediction provided by the embodiment of the present invention. As Figure 1 shown, this method may include:

[0047] S101: Measure the parameters of the permanent magnet synchronous motor at time, and the parameters at least include voltage , current and rotational speed ;

[0048] S102: Obtain the estimated value of the load torque in the prediction model through a reduced-order Luenberger load observer;

[0049] Specifically, it includes:

[0050] Based on the kinematic equations of the motor, the dynamic state-space expression is obtained:

[0051] (1)

[0052] where

[0053] ; (2)

[0054] The Luenberger reduced-order observer is designed for the above system:

[0055] (3)

[0056] where ;

[0057] The observation error is taken as:

[0058] (4)

[0059] The characteristic polynomial is written as:

[0060] (5)

[0061] The gain matrix is set, and the desired poles are configured. The desired characteristic polynomial of the observer is:

[0062] (6)

[0063] By comparing the above two characteristic polynomials, the parameters of the gain matrix can be solved: ; (7)

[0064] Assume , and at the same time, the two desired poles are equal , then the parameters of the gain matrix can be simplified as: ; (8)

[0065] By substituting the gain matrix into the state-space equation, the following can be obtained:

[0066] (9)

[0067] The first-order Euler approximation discrete form of equation (9) is:

[0068] (10)

[0069] According to the expected characteristics of the system, select the location of the poles, and construct an observer according to Equation (9) or Equation (10), then the value of the load torque can be observed.

[0070] S103: Obtain the estimated value of the moment of inertia in the prediction model through the model reference moment of inertia identification algorithm ;

[0071] Specifically, it includes:

[0072] When the sampling interval of the identification algorithm for speed is small enough:

[0073] (11)

[0074] Equation of mechanical angular velocity:

[0075] (12)

[0076] Ts is the control period of the identification algorithm. Ignoring the dynamic friction coefficient Bm, discretize and simplify the motor motion equation (12):

[0077] (13)

[0078] Assume that the motor load torque is constant. During the control period, the load torque hardly changes. , take the difference between the above two equations to get:

[0079] (14)

[0080] Define parameters for subsequent representation: , ;

[0081] According to the parameter definition, define the reference model and the adjustable model from Equation (14) respectively as:

[0082] (15)

[0083] Define the speed estimation error: .

[0084] Through the identification iteration algorithm based on MRAS, we can get:

[0085] (16)

[0086] Through the above Equation (16), obtain the estimated value of the moment of inertia .

[0087] S104: The estimated value of the load torque , the estimated value of the uncertain moment of inertia Feed it back to the prediction model to achieve the correction of model uncertainty, and use the estimated value of the load torque as the reference value of the electromagnetic torque;

[0088] S105: Obtain the current values at +1 and +2 moments through the corrected motor discrete model prediction equation and one-step delay compensation , and then obtain the predicted values of the rotational speed, magnetic flux, and electromagnetic torque ;

[0089] Specifically, the one-step delay compensation to obtain the current values at +1 and +2 moments includes:

[0090] After the generalized motor mechanical equation in the d-q coordinate system is approximated by the first-order forward Euler method, the prediction equation is obtained:

[0091] (17)

[0092] (18)

[0093] (19)

[0094] (20)

[0095] Perform one-step delay compensation:

[0096] (21)

[0097] (22)

[0098] (23)

[0099] (24).

[0100] Perform parameter perturbation on the moment of inertia and the electromagnetic torque . When the parameters do not match, the current speed prediction model can be expressed as:

[0101] (25)

[0102] where ΔJ and ΔTe represent the errors between the parameter values and the true values. Then, the prediction error between the parameter-free error model and the parameter-error model under parameter variation is:

[0103] (26).

[0104] S106: Input the predicted values of the rotational speed, magnetic flux, and electromagnetic torque, as well as the given rotational speed, magnetic flux, and electromagnetic torque into the cost function. The cost function mainly consists of three parts: the rotational speed error PI cost item , the electromagnetic torque and magnetic flux cost items , and the compensation factor : ; Among them, the rotational speed error PI cost item While controlling the rotational speed, it can eliminate the steady-state error of the speed control caused by the uncertainty of the permanent magnet flux linkage and inductance parameters. Together with the finite set of voltage vectors Perform optimal voltage vector selection to determine the optimal voltage vector, and the optimal voltage vector minimizes the cost function;

[0105] Specifically, the cost function includes a rotational speed cost function item, and the rotational speed cost function item is expressed as:

[0106] (27)

[0107] Where the speed error Minimize in the cost function, and select the optimal voltage vector to approach the predefined speed reference .

[0108] The cost function includes a magnetic flux cost function item and an electromagnetic torque cost function item, and the magnetic flux cost function item and the electromagnetic torque cost function item are respectively expressed as:

[0109] (28)

[0110] (29)

[0111] Where is the stator magnetic flux error, is the electromagnetic torque error.

[0112] The cost function item regarding speed in the initial cost function is:

[0113] (30)

[0114] The motor speed tracking error in PI form can be expressed as:

[0115] (31)

[0116] The discrete form of the above formula can be written as:

[0117] (32)

[0118] When When there is no PI cost item, the speed cost function is the basic speed error item. Therefore, the speed error item in the cost function can be expressed by the following formula:

[0119] (33)

[0120] According to the relative magnitude of the rotational speed error, it can be determined whether to use the PI cost function. The PI cost function is activated within the specified speed range, and m can be set by itself according to the desired effect:

[0121] (34)

[0122] The total cost function is:

[0123] (35).

[0124] S107: Output the optimal voltage vector to the inverter to drive the motor to operate.

[0125] The speed control method for permanent magnet synchronous motor model prediction provided by the embodiments of this application obtains an accurate prediction model by real-time updating the values of load torque and moment of inertia, and realizes the correction of the motor prediction model; when the inductance or permanent magnet flux linkage parameters do not match, a large steady-state error will be generated for speed control. By adjusting the integral coefficient in the rotational speed error PI cost item, the purpose of eliminating the speed steady-state error is achieved.

[0126] The method provided by this application will be introduced in detail through specific examples below.

[0127] Parameter definition and description: is the sampling period; , , and are the voltages applied to the magnetic flux linkage, current, and magnetic flux on the stator dq axes; is the electrical angular velocity of the motor, is the mechanical angular velocity of the motor, and ; are the electromagnetic torque and load torque of the motor respectively; is the number of pole pairs, is the magnetic flux linkage of the permanent magnet, is the moment of inertia, is the viscous friction coefficient, is the stator resistance.

[0128] In the prediction equation represents the sampling value at the current moment, represents the predicted value at the next moment, represents the predicted value after considering one-step delay compensation, and so on; the subscript Represents the parameter vector, which respectively contains the parameter values of the d-axis and q-axis, such as the current vector ; Indicates that there is an error between the nominal parameter value and the actual value of the moment of inertia and the electromagnetic torque; Is the predicted motor speed with parameter mismatch. Is the integral regulation coefficient, Is the limit value of the relative speed error, Is the speed prediction error, and the subscript Represents the parameter reference value, Represents the error between the predicted value and the reference value.

[0129] The control block diagram of the direct speed control of the permanent magnet synchronous motor is as shown in Figure 2 The figure. This control block diagram mainly includes: measurement of the parameters of the permanent magnet synchronous motor, correction of the prediction model (mainly including the load observer and the moment of inertia identification algorithm), prediction considering one-step delay compensation, cost function including the flux torque error term, speed error PI term, prediction compensation factor, reference parameter setting (speed flux and torque setting), and drive output of the three-phase inverter.

[0130] Control process: Measure the voltage, current, speed and other parameters of the permanent magnet synchronous motor through sensors. Obtain the uncertain load torque, moment of inertia and other parameters and the given electromagnetic torque in the prediction model through the load observer and the model reference moment of inertia identification algorithm. Obtain the current value at time through the model prediction equation and one-step delay compensation, and then obtain the speed, flux and electromagnetic torque. Input the predicted and given speed, flux and electromagnetic torque into the cost function, substitute the 8 voltage vectors, find the vector that minimizes the cost function, and output it to the inverter to drive the motor to operate, eliminating the influence of parameter uncertainty on the control effect while realizing the direct control of the motor speed.

[0131] Motor equation

[0132] Generalized mechanical equation of the permanent magnet synchronous motor in the d-q coordinate system:

[0133]

[0134]

[0135]

[0136]

[0137] Model correction

[0138] When the permanent magnet synchronous motor drive system is applied in many occasions, the change of the load moment of inertia is often relatively large. In the driving occasions where the load changes, it is required that the speed servo system for driving can ensure good response performance when a large external load disturbance occurs.

[0139] The present invention adopts a method of first identifying the disturbance and then compensating to solve the uncertainty of the model. The Landau algorithm is used to identify the moment of inertia, and the reduced-order Luenberger observer is used to observe the load torque. The principles of them are introduced respectively below.

[0140] In order to verify the anti-interference ability of the algorithm, the following settings are given in the simulation experiment: at 0.1 s, the load torque jumps from 2 N·m to 5 N·m; at 0.2 s, the rotational speed jumps from 500 rpm to 100 rpm.

[0141] Load torque observation:

[0142] It can be seen from the motor speed prediction model that the actual load torque of the motor is needed to predict the motor speed. In the traditional double-loop model predictive control, an outer speed control loop with a PI controller is usually used to generate the torque reference. The obtained torque reference can reflect and track the change of the load torque. In the proposed MPDSC strategy, since there is no outer speed loop with a PI controller, the estimation of the load torque becomes a problem. Therefore, an online observer based on the reduced-order Luenberger load torque is adopted to observe and track the load torque in real time.

[0143] From the motor kinematic equation, the dynamic state space expression is obtained:

[0144] (1)

[0145] Among them,

[0146] . (2)

[0147] Design a Luenberger reduced-order observer for the above system:

[0148] (3)

[0149] Among them, .

[0150] Take the observation error:

[0151] (4)

[0152] Write out the characteristic polynomial:

[0153] (5)

[0154] Similarly, by setting an appropriate gain matrix , configuring the desired poles , the desired characteristic polynomial of the observer is:

[0155] 6 (6)

[0156] By comparing the above two characteristic polynomials, the parameters of the gain matrix can be solved: (7)

[0157] Assume , and at the same time, the two desired poles are equal , then the parameters of the gain matrix can be simplified to: (8)

[0158] By substituting the gain matrix into the state - space equation, we can get:

[0159] (9)

[0160] Perform the first - order Euler approximation discrete form on equation (9):

[0161] (10)

[0162] According to the desired characteristics of the system, select the position of the poles, and construct the observer according to equation (9) or (10), then the value of the load torque can be observed. The observation effect diagram of the load torque observer is as Figure 7 shown, and a good observation effect can be achieved.

[0163] In the motor control system, the use of a load torque observer can reduce the manufacturing cost and mechanical complexity of the motor system. In addition, the use of an online load torque observer can enable the MPDSC to quickly and accurately track and control the speed change.

[0164] Moment of inertia identification:

[0165] The Model Reference Adaptive Control (MRAS) algorithm was proposed by the French scholar Landau, so it is also called the Landau algorithm. The MRAS algorithm has a recursive structure, that is, the new value of the estimated parameter is equal to the previous value plus a correction term depending on the previous measurement.

[0166] In the parameter estimation of a PMSM (Permanent Magnet Synchronous Motor) based on the MRAS (Model Reference Adaptive System) algorithm, the PMSM itself can be regarded as a reference model. Then, an adjustable model is constructed according to the mathematical model of the PMSM. Based on the difference between the outputs of the reference model and the adjustable model, the adjustable model is adjusted to minimize the output error as much as possible, so as to obtain an adjustable model that is as close as possible to the reference model. Thus, the parameters of the motor can be obtained from the adjustable model, and then the moment of inertia of the PMSM motor can be calculated. The moment of inertia can be obtained by using the MRAS discrete-time recursive algorithm.

[0167] When the sampling interval of the identification algorithm for speed is small enough:

[0168] (11)

[0169] Equation of mechanical angular velocity:

[0170] (12)

[0171] Let Ts be the control period of the identification algorithm. Ignoring the viscous friction coefficient Bm, the motor motion equation (12) is discretized and simplified:

[0172] (13)

[0173] Assume that the motor load torque is constant. During the control period, the load torque hardly changes. , Subtracting the above two equations gives:

[0174] (14)

[0175] Define parameters for subsequent representation: , .

[0176] According to the parameter definitions, the reference model and the adjustable model are defined by equation (14) respectively as:

[0177] (15)

[0178] Define the speed estimation error: .

[0179] Through the MRAS-based identification iterative algorithm, we can obtain:

[0180] (16)

[0181] From the above equation (16), the estimated value of the moment of inertia: .

[0182] The effect diagram of moment of inertia identification is as Figure 8As shown, it can be seen from the figure that the identification result is good and the identification speed is fast.

[0183] In summary, through the load torque observer and the moment of inertia identification algorithm, the uncertain parameters in the prediction model are corrected to reduce the error caused by the uncertainty of the model parameters to the prediction.

[0184] Prediction equation

[0185] After the generalized motor mechanical equation in the d-q coordinate system is approximated by the first-order forward Euler method, the prediction equation is obtained:

[0186] (17)

[0187] (18)

[0188] (19)

[0189] (20)

[0190] Since the MPDSC controller takes a certain amount of time to perform calculations, the optimal voltage vector cannot be applied to the PMSM immediately after sampling. As a result, there will be a one-step delay, which is bound to occur. If no compensation is made for it, the performance of the controller will deteriorate further. Therefore, it is necessary to compensate for the one-step delay.

[0191] (21)

[0192] (22)

[0193] (23)

[0194] (24)

[0195] Considering the one-step delay compensation, according to the current state of the machine and the possible control actions, using the current measurement values, a one-step prediction is performed according to (17)-(20) to obtain the predicted values (21)-(24) at time k+1, and these values are used as the measurement values at the next moment to perform the next prediction, and the torque, flux linkage and speed at time k+2 in the future can be predicted. The prediction schematic diagram considering the one-step delay compensation is as Figure 3 shown. Then, according to the principle of minimizing the cost function based on the speed, flux linkage and torque errors at time k+2, the optimal voltage vector to be applied at time k+1 can be determined.

[0196] Parameter sensitivity

[0197] First, analyze the parameter sensitivity of MPDSC. Based on the prediction model (20), as can be seen from (20), if the viscous coefficient is not considered in predicting the motor speed , the motor speed is related to the moment of inertia , and the electromagnetic torque .

[0198] Therefore, parameter perturbations are performed on these two variables. When the parameters do not match, the current speed prediction model can be expressed as:

[0199] (25)

[0200] Where ΔJ and ΔTe represent the errors between the parameter values and the true values. Then the prediction errors of the parameterless error model and the model with parameter errors under parameter variations are:

[0201] (26)

[0202] As can be seen from the above equation, the mismatch or uncertainty of any parameter will lead to errors in the predicted speed. Figure 5 、 Figure 6 show the relationship between the prediction error and the parameter mismatch. It can be seen that the uncertainties of the moment of inertia and the electromagnetic torque parameters will produce prediction errors in speed prediction. Among the motor parameters, the stator resistance has little influence on the prediction effect, while the inductance , the permanent magnet flux linkage have a significant impact on the predicted current, and thus have an impact on the flux, torque, and speed. When their parameters do not match, a steady-state error in speed control will be generated. Therefore, correctly identifying the moment of inertia , dealing with the inductance , the permanent magnet flux linkage and other parameters to bring the steady-state error is of great significance for improving the speed prediction effect.

[0203] Basic cost function:

[0204] Considering the compensation of one-step delay, the form of the cost function is given. Different from the traditional MPC strategy, the main control objective of the proposed MPDSC strategy is the motor speed. By integrating the outer speed loop into the MPC strategy, this cascaded linear structure with a PI controller can be eliminated.

[0205] Therefore, the main consideration in the cost function is the speed error error, and the speed cost function term is expressed as:

[0206] (27)

[0207] Speed error Minimized in the cost function, so the optimal voltage vector is selected to approach the predefined speed reference . Since the speed error is directly used in the control algorithm , the use of a PI controller in the control loop is avoided. Therefore, the tuning process of PI parameters is eliminated.

[0208] The speed error included in the cost function Contributes to the speed regulation, so the stator current needs to be controlled separately. Since the quality of the stator current can be indirectly controlled by the of the stator flux linkage, the of the stator flux linkage error is added to the cost function to improve the quality of the stator current. The flux cost function is expressed as:

[0209] (28)

[0210] The above two equations respectively define the two main elements in the cost function: motor speed and flux. The speed reference can be flexibly adjusted according to the requirements of actual applications. The flux reference is usually determined according to the maximum torque per ampere to achieve high efficiency.

[0211] To control the electromagnetic torque, the electromagnetic torque needs to be limited. Since the speed outer loop has been cancelled, the reference value of the torque cannot be obtained. However, when operating at a constant speed, the electromagnetic torque is equal to the load torque, so the load torque can be observed and the observed value is set equal to the reference value of the electromagnetic torque. The electromagnetic torque error is added to the cost function :

[0212] (29)

[0213] After forming the basic cost function including the rotational speed, electromagnetic torque and flux terms, the speed tracking ability, current torque ripple and speed overshoot are enhanced by introducing prediction correction, load torque estimation and rotational speed PI cost function selection items.

[0214] Prediction compensation factor:

[0215] If the performance after the k+2 moment is not considered, the minimum speed error can be achieved by greatly changing the speed, which is a process of local optimization operation, that is, the selected voltage vector is only optimal at the k + 2 moment.

[0216] However, if the selected voltage vector is not globally optimal during a relatively long control time, it may lead to large torque and speed ripples.

[0217] This is because the predicted number of steps is too short, which may affect the transient and steady-state control performance of the control system. During startup or speed change, too short a prediction will cause large overshoot and oscillation. During steady-state operation, without considering the compensation for too short a prediction step, the flux and torque fluctuations are large. Therefore, a simple and effective method proposed by Ming Liu et al. is adopted to solve this problem and improve the prediction effect. The compensation factor f is defined as:

[0218] (30)

[0219] Considering the complexity of the multi-step prediction calculation amount, based on predict 2 steps backward, and only calculate the prediction errors at the th and the th moments. The magnitude of the weight coefficient reflects the influence of different moments. The closer the future moment is to the current moment, the greater its influence factor on the prediction. The weight coefficients of the prediction errors at the two moments are set to 1 / 2 and 1 / 3 respectively.

[0220] PI cost function:

[0221] In model predictive control, the mathematical model of the system is used to predict the values of state variables. There are problems of parameter mismatch and model uncertainty in the control performance of model predictive control. Due to inaccurate prediction, there will be a steady-state error. When the permanent magnet flux linkage and inductance values change, the speed of MPDSC control will generate a steady-state error. To improve the robustness of model predictive control, a simple predictive speed control strategy is proposed. By designing a cost function in the form of proportional integral (PI), which acts on the speed cost term, the integral action is only activated within a predetermined range, and the cumulative error is weighted by the sampling time, which is beneficial to the design of the integral coefficient.

[0222] The cost function term regarding speed in the initial cost function is:

[0223] The motor speed tracking error in the form of PI can be expressed as:

[0224] (31)

[0225] The discrete form of the above formula can be written as:

[0226] (32)

[0227] When , it is equivalent to having no PI cost term, and the speed cost function is the basic speed error term. Therefore, the speed error term in the cost function can be expressed by the following formula:

[0228] (33)

[0229] According to the relative magnitude of the rotational speed error, it can be determined whether to use the PI cost function. The PI cost function is activated within the specified speed range, and m can be set by oneself according to the desired effect:

[0230] (34)

[0231] To sum up, after considering the one-step delay compensation, the total cost function of MPDSC is:

[0232] (35)

[0233] To sum up, by constructing a cost function including the PI term of the motor speed, the flux term, the torque term, and the prediction compensation term, while controlling the speed, torque, etc. of the motor, the current ripple, torque ripple are reduced, and the steady-state error caused by the uncertainty of the inductance and permanent magnet flux linkage parameters is eliminated. The control flow chart of the proposed direct speed control is as Figure 4 shown.

[0234] Parameter Sensitivity Simulation Experiment

[0235] Considering the parameter uncertainty of MPDSC, the load torque observer is applied to observe the load torque, and the MRAS algorithm is applied to identify the moment of inertia to improve the parameter robustness. At the same time, the PI cost function is applied to solve the speed steady-state error generated by MPDSC when the inductance and and as well as the permanent magnet flux linkage are different from the nominal values.

[0236] The simulation parameters are as follows: number of pole pairs , permanent magnet flux linkage , d-q axis inductance , moment of inertia , viscous coefficient , sampling time , DC bus voltage , reference flux linkage , reference speed , reference speed rad / s.

[0237] Simulation Experiment:

[0238] In the following simulation experiments, the load torque is given , and the speed is given . Under this condition, other motor parameters are changed.

[0239] 1. The mismatch of the stator inductance parameters will affect the prediction effect. By changing the inductance of the motor parameters to , , the original MPDSC control has a steady-state error. The simulation result diagram is shown in Fig. 9(a). The control effect diagram of the improved MPDSC with the speed PI cost function is shown in Fig. 9(b). It can be stabilized to the given speed, and the steady-state error and overshoot of the speed are reduced.

[0240] 2. The permanent magnet flux linkage has a great influence on the speed steady-state error. When the permanent magnet flux linkage parameters are not mismatched, the speed steady-state error is very small. The simulation result diagrams are shown in Fig. 10(a) and Fig. 10(b). When it is less than the original magnetic flux linkage, the speed steady-state value is lower than the given value. When it is greater than the original magnetic flux linkage, the speed steady-state value is higher than the given value.

[0241] Fix the load torque and speed, and only change the permanent magnet flux linkage to . If the PI cost function is not adopted, the MPDSC control cannot be stabilized around the reference speed and will generate a steady-state error. After adopting the cost function improvement, the steady-state error can be reduced and the control converges to the reference speed.

[0242] 3. The moment of inertia will affect the control speed of MPDSC. When the moment of inertia is inconsistent with the nominal value, the simulation realizes the parameter uncertainty by changing the value of the moment of inertia to . The simulation effect diagrams of the unimproved and improved MPDSC are shown in Fig. 11(a) and Fig. 11(b). It can be seen from Fig. 11(a) and Fig. 11(b) that through the improvement, the value of the moment of inertia is accurately identified. Although the adjustment time becomes longer, the overshoot of the electromagnetic torque and stator current becomes smaller, and the electromagnetic torque curve is smoother.

[0243] In summary, the present invention proposes a model predictive direct speed control and its parameter robustness control method. By considering the prediction error problem caused by parameter uncertainty, based on the load observer and the moment of inertia identification algorithm, the accuracy of the model is corrected. At the same time, by introducing the motor PI cost term into the cost function, the elimination of the steady-state error in the case of parameter mismatches such as inductance and permanent magnet flux linkage is realized, and finally the parameter robustness control is achieved.

[0244] See Figure 12 , corresponding to a speed control method for a permanent magnet synchronous motor model prediction provided in an embodiment of the present application. As Figure 12 shown, an embodiment of the present application also provides a speed control system for a permanent magnet synchronous motor model prediction. The system may specifically include:

[0245] A reference acquisition unit 201, configured to measure the permanent magnet synchronous motor through a sensor Parameters at a moment, where the parameters at least include voltage, current, and rotational speed;

[0246] A load torque estimation unit 202, configured to obtain an estimated value of the load torque in a prediction model through a reduced-order Luenberger load observer;

[0247] A moment of inertia estimation unit 203, configured to obtain an estimated value of the moment of inertia in a prediction model through a model reference moment of inertia identification algorithm;

[0248] A model correction unit 204, configured to feedback the estimated value of the load torque and the estimated value of the uncertain moment of inertia back to the prediction model to achieve correction of model uncertainty, and use the estimated value of the load torque as a reference value of the electromagnetic torque;

[0249] A compensation unit 205, configured to obtain +1, the current values at +2 moments through a corrected motor discrete model prediction equation and one-step delay compensation, and further obtain predicted values of rotational speed, magnetic flux, and electromagnetic torque;

[0250] An optimal voltage vector selection unit 206, configured to input the predicted values of the rotational speed, magnetic flux, and electromagnetic torque, as well as the given rotational speed, magnetic flux, and electromagnetic torque, into a cost function; and perform optimal voltage vector selection together with a finite set of voltage vectors to determine an optimal voltage vector, where the optimal voltage vector minimizes the cost function;

[0251] An optimal voltage vector output unit 207, configured to output the optimal voltage vector to an inverter to drive the motor to operate.

[0252] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the existence of additional identical elements in the process, method, article or device comprising the element.

[0253] From the description of the above embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of this application, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.

[0254] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for a system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple. For the relevant parts, reference can be made to the partial description of the method embodiment. The systems and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0255] The above are only the preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention are all included in the protection scope of the present invention.

Claims

1. A speed control method for a permanent magnet synchronous motor based on model prediction, characterized in that, The method includes: Measuring the parameters of the permanent magnet synchronous motor at time k through sensors, where the parameters at least include voltage, current, and rotational speed; Obtaining the estimated value of the load torque in the prediction model through a reduced-order Luenberger load observer; Obtaining the estimated value of the uncertain moment of inertia in the prediction model through a model reference moment of inertia identification algorithm, including: When the sampling interval of the speed by the identification algorithm is small enough: dω m / dt≈(ω m (k)-ω m (k - 1)) / Δt=(ω m (k)-ω m (k - 1)) / T s (11) Equation of mechanical angular velocity: T s For the identification algorithm control period, equation (12) is discretely simplified as follows: Assume that the load torque of the motor is constant. During the control period, the load torque remains unchanged. Taking the difference between Equation (12) and Equation (13) gives: Define the parameters: b(k) = T s / J, U(k - 1) = T e (k - 1) - T e (k - 2); According to the parameter definition, the reference model and the adjustable model are respectively defined by equation (14) as: Define the speed estimation error: Obtained through an identification iterative algorithm based on MRAS: The estimated value of the moment of inertia is obtained through the above formula (16). Feeding back the estimated value of the load torque and the estimated value of the uncertain moment of inertia to the prediction model to correct the model uncertainty, and using the estimated value of the load torque as the reference value of the electromagnetic torque; Obtaining the current values at times k + 1 and k + 2 through the corrected motor discrete model prediction equation and one-step delay compensation, and further obtaining the predicted values of rotational speed, magnetic flux, and electromagnetic torque; the one-step delay compensation for obtaining the current values at times k + 1 and k + 2 includes: After the generalized motor mechanical equation in the d-q coordinate system is approximated by the first-order forward Euler method, the prediction equation is obtained: Compensating with one-step delay: Inputting the predicted values of the rotational speed, magnetic flux, and electromagnetic torque and the given rotational speed, magnetic flux, and electromagnetic torque into the cost function; together with the finite set of voltage vectors, an optimal voltage vector is selected to determine the optimal voltage vector, and the optimal voltage vector minimizes the cost function; Outputting the optimal voltage vector to the inverter to drive the motor to operate.

2. The speed control method for a permanent magnet synchronous motor based on model prediction according to claim 1, characterized in that, The obtaining of the estimated value of the uncertain load torque in the prediction model through the reduced-order Luenberger load observer includes: Obtaining the dynamic state space expression from the motor kinematic equation: Where, Designing a Luenberger reduced-order observer for the above system: Among them, Taking the observation error: Writing out the characteristic polynomial: Set the gain matrix Configure the desired poles α, β, and the desired characteristic polynomial of the observer: s 2 -(α + β)s + αβ = 0 (6) By comparing the characteristic polynomial (5) and the polynomial (6), the parameters of the gain matrix are solved: Assume B m = 0. Meanwhile, if the two desired poles are equal, i.e., α = β, the parameters of the gain matrix are simplified to: By substituting the gain matrix into the state space equation, we get: Performing a first-order Euler approximation discrete form on equation (9): According to the desired characteristics of the system, the position of the poles is selected, and an observer is constructed according to equation (9) or (10) to observe the estimated value of the load torque.

3. The speed control method for permanent magnet synchronous motor model prediction according to claim 1, characterized in that, For the moment of inertia J and the electromagnetic torque T e perform parameter perturbation. When the parameters do not match, the current speed prediction model is expressed as: Where, ΔJ and ΔTe represent the errors between the parameter values and the true values, then the prediction error between the parameterless error model and the model with parameter errors under parameter changes is:

4. The speed control method for permanent magnet synchronous motor model prediction according to claim 1, characterized in that, The cost function includes a rotational speed cost function term, and the rotational speed cost function term is expressed as: ω error = (ω - ω ref ) 2 (27) Among them, the speed error ω error is minimized in the cost function, and the optimal voltage vector is selected to approach the predefined speed reference ω ref .

5. The speed control method for permanent magnet synchronous motor model prediction according to claim 4, characterized in that, The cost function includes a magnetic flux cost function term and an electromagnetic torque cost function term, and the magnetic flux cost function term and the electromagnetic torque cost function term are respectively expressed as: T error = (T e - T ref ) 2 (29) wherein, is the stator flux linkage error, and T error is the electromagnetic torque error.

6. The speed control method for permanent magnet synchronous motor model prediction according to claim 5, characterized in that, The cost function term regarding speed in the initial cost function is: e k+1 = ω ref - ω e k+1 (30) The PI-form motor rotational speed tracking error is expressed as: The discrete form of the above formula is written as: S k = S k-1 +(e k - e k-1 ) + K i ·e k (k)·T s (32) When K i = 0, the speed error term in the cost function is expressed by the following equation: Activating the PI cost function within the specified speed range, where m is set by oneself according to the desired effect: The total cost function is:

7. A speed control system for a permanent magnet synchronous motor based on model prediction, characterized in that, For implementing the speed control method for permanent magnet synchronous motor model prediction according to any one of claims 1-6, the system includes: A reference acquisition unit for measuring the parameters of the permanent magnet synchronous motor at time k through sensors, where the parameters at least include voltage, current, and rotational speed; A load torque estimation unit for obtaining an estimated value of the load torque in the prediction model through a reduced-order Luenberger load observer; A moment of inertia estimation unit for obtaining an estimated value of the uncertain moment of inertia in the prediction model through a model reference moment of inertia identification algorithm; A model correction unit for feeding back the load torque estimated value and the uncertain moment of inertia estimated value to the prediction model to achieve correction of model uncertainty and using the load torque estimated value as a reference value for the electromagnetic torque; A compensation unit for obtaining current values at times k + 1 and k + 2 through the corrected motor discrete model prediction equation and one-step delay compensation, and further obtaining predicted values of the rotational speed, magnetic flux, and electromagnetic torque; An optimal voltage vector selection unit for inputting the predicted values of the rotational speed, magnetic flux, and electromagnetic torque and the given rotational speed, magnetic flux, and electromagnetic torque into a cost function; and performing optimal voltage vector selection together with a finite set of voltage vectors to determine an optimal voltage vector, where the optimal voltage vector minimizes the cost function; An optimal voltage vector output unit for outputting the optimal voltage vector to an inverter to drive the motor to operate.

Images