A method and device for speed control of a permanent magnet synchronous motor

Abstract

The application provides a permanent magnet synchronous motor speed control method and device. The method comprises: obtaining a mathematical model, a reference speed, an actual speed and an actual q-axis current; generating a basic q-axis reference current according to a speed error and constructing a finite candidate prediction correction amount set; online identifying a rotational inertia and a viscous friction coefficient, updating a mechanical motion equation parameter of a disturbance observer; estimating a load disturbance based on the updated mechanical motion equation, generating a disturbance compensation current, and introducing a load disturbance estimation value as an equivalent disturbance item into a prediction model; rolling prediction is performed on each candidate prediction correction amount to obtain a future speed and a future q-axis current; the optimal prediction correction amount is determined according to the future speed, the future q-axis current, a prediction voltage and a dynamic weight cost function; and the final q-axis reference current is fused and generated, and is used as a q-axis current instruction of a current inner loop. The application can improve the speed tracking performance and the anti-disturbance ability.

Description

Technical Field

[0001] This application relates to the field of permanent magnet synchronous motor control technology, and in particular to a closed-loop speed control technology for permanent magnet synchronous motors that combines disturbance observation, online parameter identification, and rolling optimization of finite candidate sets. Background Technology

[0002] Permanent magnet synchronous motors (PMSMs) have advantages such as compact structure, high power density, high efficiency, and fast dynamic response, and are widely used in electric vehicles, robots, CNC machine tools, servo drives, and industrial automation. For PMSM speed control systems, the controller not only needs to maintain good speed tracking performance when the reference speed changes, but also needs to suppress speed drops and shorten system recovery time under sudden load changes, model parameter variations, or external disturbances.

[0003] Currently, proportional-integral (PI) control is commonly used in the speed loop control of permanent magnet synchronous motors. PI control is simple in structure and easy to implement in engineering, but its control effect usually depends on fixed control parameters. When the load torque changes suddenly or the motor parameters deviate, the speed proportional-integral (PI) controller often needs to rely on the accumulation of speed error before adjusting the q-axis reference current, which can easily lead to problems such as response lag, large speed drops, and long recovery times, making it difficult to balance fast dynamic response and disturbance rejection performance.

[0004] Model Predictive Control (MPC) can comprehensively consider future speed errors, control input variations, and constraints such as current and voltage within a finite prediction time domain, making it suitable for optimizing the speed control process of permanent magnet synchronous motors (PMSMs). However, directly solving continuous optimization problems involves a large amount of online computation, requiring high real-time computing power from the controller and hindering rapid execution of the speed loop. Therefore, how to reduce the amount of online optimization computation while maintaining the predictive control's ability to handle constraints and dynamic responses is a problem that needs to be solved in the speed predictive control of PMSMs.

[0005] A disturbance observer (DOB) can estimate load disturbances and improve the system's immunity through disturbance compensation. In existing control schemes, the disturbance observer is typically used primarily as a feedforward compensation element, generating compensating control variables based on the estimated load disturbance values. However, during candidate control variable evaluation or predictive model updates, the estimated load disturbance values ​​may not be fully utilized, making it difficult for the predictive model to reflect the impact of load disturbances on future speed changes in a timely manner. Furthermore, the mechanical motion equations in the disturbance observer rely on model parameters such as moment of inertia and viscous friction coefficient; deviations between actual and nominal parameters can easily affect the accuracy of load disturbance estimation.

[0006] Furthermore, the cost function weights in model predictive control are typically obtained through offline tuning and remain fixed during operation. Fixed weights make it difficult to adjust the control objective in real time based on load disturbances, changes in model parameters, or changes in prediction errors. When the error between the predictive model and the actual system increases, if fixed weights are still used, the controller may not be able to enhance its speed tracking capability in a timely manner; conversely, if the weights are set too large, it may lead to increased fluctuations in the q-axis reference current, affecting the system's operational stability.

[0007] Therefore, it is necessary to provide a speed control method for permanent magnet synchronous motors to improve speed tracking performance, disturbance suppression capability, and operational stability under conditions of sudden load changes and model parameter variations, while ensuring that the online computation is controllable. Summary of the Invention

[0008] This application provides a speed control method and apparatus for permanent magnet synchronous motors, which addresses the problems in existing speed control methods for permanent magnet synchronous motors, such as insufficient disturbance rejection capability of proportional-integral control, large online computational load of continuous model predictive control, insufficient adaptability of fixed-weight model predictive control to changes in operating conditions, and insufficient participation of disturbance observers in the prediction optimization process.

[0009] On the one hand, this application provides a speed control method for a permanent magnet synchronous motor, including: Obtain the mathematical model of the permanent magnet synchronous motor, as well as the reference speed, actual speed, and actual q-axis current at the current sampling moment; Based on the speed error between the reference rotational speed and the actual rotational speed, a basic q-axis reference current is generated, and a finite set of candidate prediction corrections is constructed. Based on the speed data and q-axis current data during the operation of the permanent magnet synchronous motor, the estimated values ​​of the moment of inertia and the viscous friction coefficient are identified online, and the mechanical motion equation parameters used by the disturbance observer are updated based on the estimated values ​​of the moment of inertia and the viscous friction coefficient. Based on the updated mechanical motion equations, the actual rotational speed, and the actual q-axis current, the load disturbance of the permanent magnet synchronous motor is estimated using the disturbance observer to obtain a load disturbance estimate. A disturbance compensation current is generated based on the load disturbance estimate, and the load disturbance estimate is used as an equivalent disturbance term and introduced into the prediction model established based on the mathematical model of the permanent magnet synchronous motor. For each candidate prediction correction in the finite candidate prediction correction set, a test q-axis reference current is constructed based on the basic q-axis reference current and the corresponding candidate prediction correction. The actual rotational speed and the actual q-axis current are used as the initial prediction values. Based on the prediction model after introducing the equivalent disturbance term, a finite-step forward rolling prediction is performed to obtain the future rotational speed and future q-axis current in the prediction time domain. The speed prediction error is determined based on the future speed and the reference speed of the corresponding prediction step, and the weight of the cost function is dynamically adjusted based on the variance of the speed prediction error to obtain the dynamic weight cost function. The current constraint is determined based on the future q-axis current; the predicted voltage in the prediction time domain is determined based on the test q-axis reference current, the future q-axis current, and the mathematical model of the permanent magnet synchronous motor; and the voltage constraint is determined based on the predicted voltage; based on the current constraint, the voltage constraint, and the dynamic weight cost function, the optimal prediction correction amount is determined from the set of finite candidate prediction correction amounts. The base q-axis reference current, the optimal prediction correction, and the disturbance compensation current are combined to generate the final q-axis reference current, which is then used as the q-axis current command for the inner current loop to control the speed of the permanent magnet synchronous motor.

[0010] In one possible implementation, the mathematical model of the permanent magnet synchronous motor includes voltage equations, electromagnetic torque equations, and mechanical motion equations in a synchronously rotating dq coordinate system. The mechanical motion equations are used for speed prediction by the prediction model, load disturbance estimation by the disturbance observer, and online identification of estimated values ​​of rotational inertia and viscous friction coefficient.

[0011] In one possible implementation, generating the base q-axis reference current and constructing a finite set of candidate prediction corrections includes: performing a first-order filter on the original reference speed to obtain a smoothed reference speed; determining the speed error based on the difference between the smoothed reference speed and the actual speed; inputting the speed error into a speed proportional-integral controller to obtain an unlimited base q-axis reference current; performing a limiting process on the unlimited base q-axis reference current to obtain the base q-axis reference current; and generating multiple candidate prediction corrections based on a preset search interval and a preset search step size to form a finite set of candidate prediction corrections.

[0012] In one possible implementation, online identification of the estimated moment of inertia and the estimated coefficient of viscous friction includes: constructing a parameter identification regression model based on the discrete form of the mechanical motion equations; updating the parameter vector in the parameter identification regression model online using a recursive least squares algorithm with a forgetting factor; determining the estimated moment of inertia and the estimated coefficient of viscous friction based on the updated parameter vectors; and outputting the estimated moment of inertia and the estimated coefficient of viscous friction to a disturbance observer after applying upper and lower limits or a rate of change limit to the estimated values, so as to update the mechanical motion equation parameters used by the disturbance observer.

[0013] In one possible implementation, the disturbance observer, based on the updated mechanical motion equations, obtains the original load disturbance estimate according to the estimated moment of inertia, the estimated viscous friction coefficient, the mechanical angular velocity corresponding to the current actual speed, and the actual q-axis current. The original load disturbance estimate is then low-pass filtered and saturated to obtain a new load disturbance estimate. A disturbance compensation current is generated based on the load disturbance estimate and the feedforward compensation coefficient. The load disturbance estimate is also introduced as an equivalent disturbance term into the prediction model established based on the mathematical model of the permanent magnet synchronous motor, allowing the prediction model to consider the impact of load disturbance on future speed changes during the evaluation of candidate prediction corrections.

[0014] In one possible implementation, for any candidate prediction correction in the finite set of candidate prediction corrections, the base q-axis reference current is added to the candidate prediction correction to obtain the test q-axis reference current; using the actual rotational speed and actual q-axis current at the current sampling time as the initial prediction values, based on the test q-axis reference current and the prediction model after introducing the equivalent disturbance term, a finite-step forward rolling prediction is performed on the future q-axis current and future rotational speed in the prediction time domain to obtain the future rotational speed and future q-axis current in the prediction time domain corresponding to the candidate prediction correction.

[0015] In one possible implementation, dynamically adjusting the cost function weights includes: calculating the speed prediction error variance in the prediction time domain; filtering the speed prediction error variance to obtain a smoothed prediction error variance estimate; updating the weight ratio between the speed tracking error weight and the control increment penalty weight based on the smoothed prediction error variance estimate; limiting the weight ratio; and constructing a dynamic weight cost function based on the speed tracking error weight, the control increment penalty weight, and the q-axis current penalty weight.

[0016] In one possible implementation, determining the optimal prediction correction includes: judging whether the prediction results corresponding to each candidate prediction correction satisfy the current constraint and voltage constraint; determining the candidate prediction correction that does not satisfy the current constraint or voltage constraint as an infeasible solution; for the candidate prediction correction that satisfies the current constraint and voltage constraint, calculating its corresponding dynamic weight cost function value; and determining the candidate prediction correction with the smallest dynamic weight cost function value as the optimal prediction correction.

[0017] On the other hand, this application provides a speed control device for a permanent magnet synchronous motor, including a model acquisition module, a basic current and candidate quantity generation module, an online parameter identification module, a disturbance observation and compensation module, a rolling prediction module, an adaptive weight adjustment module, a constraint optimization module, and a current synthesis control module. Each module is used to execute the corresponding steps in the aforementioned permanent magnet synchronous motor speed control method.

[0018] The permanent magnet synchronous motor speed control method and device provided in this application have at least the following beneficial effects: First, this application constructs a finite set of candidate prediction corrections and performs finite-step forward rolling prediction, constraint judgment, and dynamic weight cost function evaluation on the candidate prediction corrections. This avoids directly solving complex optimization problems in continuous form, reduces the computational cost of online optimization of the velocity loop, and facilitates real-time implementation.

[0019] Second, this application identifies the estimated values ​​of rotational inertia and viscous friction coefficient online, and uses the identification results to update the mechanical motion equation parameters used by the disturbance observer, so that the disturbance observer can update the model parameters according to the motor operating state, thereby improving the accuracy of load disturbance estimation.

[0020] Third, this application uses the load disturbance estimate output by the disturbance observer to generate the disturbance compensation current and as the equivalent disturbance term in the prediction model, so that the prediction model can consider the impact of load disturbance on future speed changes during the evaluation of candidate prediction corrections, thereby improving the accuracy of the prediction results and the disturbance rejection performance under sudden load conditions.

[0021] Fourth, this application determines the speed prediction error based on the future speed in the prediction result and the reference speed of the corresponding prediction step, and dynamically adjusts the weight of the cost function according to the variance of the speed prediction error, so that the controller can adjust the optimization target as the prediction error changes, thereby enhancing the speed tracking performance while suppressing excessive fluctuations in the control quantity.

[0022] Fifth, this application integrates the basic q-axis reference current, the optimal prediction correction, and the disturbance compensation current to generate the final q-axis reference current, and uses the final q-axis reference current as the q-axis current command of the current inner loop. This enables coordinated control among the speed PI control reference, prediction correction, and disturbance compensation, thereby improving the speed tracking performance, disturbance suppression capability, and operational stability of the permanent magnet synchronous motor under conditions of sudden load changes and parameter variations. Attached Figure Description

[0023] Figure 1 A flowchart of a speed control method for a permanent magnet synchronous motor provided in this application embodiment; Figure 2 This application provides an online parameter identification, disturbance observation, and prediction model correction data flow diagram. Figure 3 A flowchart of rolling optimization of finite candidate prediction correction amount is provided for an embodiment of this application; Figure 4 A schematic diagram of a closed-loop speed control system for a permanent magnet synchronous motor is provided in this application embodiment; Figure 5 The speed response curves under different control strategies provided in the embodiments of this application; Figure 6 Electromagnetic torque response curves under different control strategies provided in the embodiments of this application; Figure 7 Load disturbance estimation curves provided for embodiments of this application; Figure 8 The q-axis disturbance compensation current curve provided in the embodiments of this application; Figure 9 This is a structural diagram of a speed control device for a permanent magnet synchronous motor provided in an embodiment of this application. Detailed Implementation

[0024] The embodiments of this application will now be described with reference to the accompanying drawings. It should be understood that the following embodiments are used to illustrate the technical solutions of this application and are not intended to limit the scope of protection of this application. Where there is no conflict, the technical features in the embodiments of this application can be combined with each other.

[0025] In this embodiment, the actual rotational speed of the permanent magnet synchronous motor can be expressed as mechanical angular velocity, or it can be calculated from mechanical angular velocity. For example, the actual rotational speed can be expressed as mechanical angular velocity. Or expressed as mechanical angular velocity The corresponding rotational speed n. For ease of description, the mechanical angular velocity will be used below. This represents the actual rotational speed of the permanent magnet synchronous motor. Correspondingly, the reference speed can be expressed as the reference mechanical angular velocity. The q-axis current refers to the q-axis current in the synchronously rotating dq coordinate system.

[0026] In this embodiment, the load disturbance can be represented as a load torque disturbance, and the load disturbance estimate can be a load torque estimate or an equivalent disturbance related to the load torque disturbance. The load disturbance estimate output by the disturbance observer is used to generate the disturbance compensation current on the one hand, and on the other hand, it participates in the rolling prediction as an equivalent disturbance term in the prediction model.

[0027] Example 1 Please see Figure 1 This application provides a speed control method for a permanent magnet synchronous motor. The method focuses on the speed closed-loop control system of a permanent magnet synchronous motor. It generates a final q-axis reference current through a combination of a basic q-axis reference current, finite candidate set rolling optimization, online parameter identification, disturbance observation compensation, and a dynamic weighted cost function. This final q-axis reference current is then used as the q-axis current command for the inner current loop to control the speed of the permanent magnet synchronous motor.

[0028] like Figure 1 As shown, the speed control method for the permanent magnet synchronous motor may include the following steps.

[0029] S101, obtain the mathematical model of the permanent magnet synchronous motor and the reference speed, actual speed and actual q-axis current at the current sampling time.

[0030] In one embodiment, the mathematical model of the permanent magnet synchronous motor includes voltage equations, electromagnetic torque equations, and mechanical motion equations in a synchronously rotating dq coordinate system. The voltage equations characterize the relationship between d-axis voltage, q-axis voltage, d-axis current, q-axis current, inductive component, stator resistance, permanent magnet flux linkage, and electric angular velocity. The electromagnetic torque equations characterize the relationship between q-axis current and electromagnetic torque. The mechanical motion equations characterize the relationship between electromagnetic torque, moment of inertia, mechanical angular velocity, viscous friction coefficient, and load disturbance.

[0031] In one implementation, the voltage equation of the permanent magnet synchronous motor in the synchronous rotating dq coordinate system can be expressed as:

[0032]

[0033] in, The voltage along the d-axis. This is the q-axis voltage. For d-axis current, For q-axis current, For stator resistance, For d-axis inductance, It is the q-axis inductance. Electric angular velocity, It is a permanent magnet flux linkage.

[0034] In one implementation, the permanent magnet synchronous motor can be a surface-mounted permanent magnet synchronous motor. For a surface-mounted permanent magnet synchronous motor, the electromagnetic torque equation is:

[0035] The equation of motion for a machine can be expressed as:

[0036] in, For electromagnetic torque, The torque coefficient, For rotational inertia, For mechanical angular velocity, The coefficient of viscous friction, This represents the load torque.

[0037] Substituting the electromagnetic torque equation into the mechanical motion equation, we can obtain the relationship between the q-axis current, mechanical angular velocity, and load disturbance:

[0038] This formula can serve as the basis for velocity prediction by the prediction model, load disturbance estimation by the disturbance observer, and online identification of estimated values ​​of moment of inertia and viscous friction coefficient.

[0039] At the current sampling time k, obtain the reference rotational speed, actual rotational speed, and actual q-axis current. The actual rotational speed can be obtained from an encoder, a rotary transformer, or a speed estimation algorithm, and the actual q-axis current can be obtained from the current sampling value through coordinate transformation.

[0040] S102, based on the speed error between the reference speed and the actual speed, generates the basic q-axis reference current and constructs a finite set of candidate prediction corrections.

[0041] In one implementation, to reduce the impact of sudden changes in the reference speed on the speed loop control output, the original reference speed can be first filtered to obtain a smoothed reference speed. The smoothed reference speed can be expressed as:

[0042] in, The original reference rotational speed at the k-th sampling time. The filtered, smoothed reference speed. The reference speed filter coefficient, and 0 ≤ ≤1.

[0043] The speed error is obtained by calculating the difference between the smoothed reference speed and the actual speed.

[0044] in, For speed error, This represents the actual rotational speed at the k-th sampling time.

[0045] The speed error is input to the speed proportional-integral controller to obtain the unlimited base q-axis reference current. The unlimited base q-axis reference current can be expressed as:

[0046] in, For the unlimited base q-axis reference current, The proportional coefficient of the speed proportional-integral controller. The integral coefficient of the speed proportional-integral controller. The sampling period is j This is the historical sampling sequence number. This is the integral term for the speed error.

[0047] The unlimited base q-axis reference current is limited to obtain the base q-axis reference current:

[0048] in, Based on the q-axis reference current, This is a saturation limiting function used to limit the input quantity within the parentheses to the preset upper and lower limits of the q-axis current.

[0049] After obtaining the basic q-axis reference current, a finite set of candidate predicted correction quantities is constructed based on a preset search interval and a preset search step size. Specifically, at the current sampling time... k The search interval and search step size for the prediction correction can be preset, and multiple candidate prediction corrections can be generated based on the search interval and search step size to form a finite set of candidate prediction corrections.

[0050] in, The current sampling time k The corresponding set of finite candidate prediction corrections, Let m be the correction amount for the m-th candidate prediction. M This represents the number of corrections to the candidate predictions.

[0051] For any candidate prediction correction in the finite set of candidate prediction corrections, a test q-axis reference current can be constructed based on the base q-axis reference current and the candidate prediction correction during subsequent rolling predictions:

[0052] in, The test q-axis reference current corresponds to the m-th candidate prediction correction.

[0053] Therefore, the basic q-axis reference current provides the baseline control quantity for speed control, and the finite set of candidate prediction correction quantities provides candidate correction quantities for the prediction model to perform online evaluation and optimization. This allows the prediction model to determine the optimal prediction correction quantity by rolling prediction, constraint judgment and cost function comparison of a finite number of candidate prediction correction quantities, without having to directly solve the continuous form optimization problem.

[0054] S103 identifies the estimated values ​​of rotational inertia and viscous friction coefficient online based on rotational speed data and q-axis current data, and updates the mechanical motion equation parameters used by the disturbance observer.

[0055] In one implementation, the online parameter identification module employs a recursive least squares algorithm with a forgetting factor to identify estimated values ​​of rotational inertia and viscous friction coefficient online based on the rotational speed data and q-axis current data during the operation of the permanent magnet synchronous motor. The rotational speed data may include actual rotational speeds at multiple sampling times, and the q-axis current data may include actual q-axis currents at multiple sampling times.

[0056] Specifically, a parameter identification regression model is constructed based on the discrete form of the mechanical motion equations of a permanent magnet synchronous motor. The online parameter identification module employs a recursive least squares algorithm with a forgetting factor, using historical speed data and q-axis current data from the permanent magnet synchronous motor's operation to identify estimated values ​​for the moment of inertia and viscous friction coefficient online. From the discrete form of the mechanical motion equations, we can obtain:

[0057] in, This represents the actual rotational speed at the previous sampling time. This is the estimated load disturbance value at the previous sampling time.

[0058] The discrete form of the mechanical motion equations is transformed into a parameter identification form, and a regression model is constructed:

[0059] in:

[0060]

[0061]

[0062] In the formula, Represents the regression vector The transpose of the model. The output in the regression model is determined by the change in rotational speed at adjacent sampling times. The regression vector consists of the q-axis current, mechanical angular velocity, and the estimated load disturbance value at the previous sampling time. The parameter vector contains parameters to be identified related to the moment of inertia and the coefficient of viscous friction.

[0063] The parameter vector is updated using a recursive least squares algorithm with a forgetting factor:

[0064]

[0065]

[0066] in, For recursive least squares gain, Let covariance matrix be the variance matrix. It is the forgetting factor, and 0 < ≤1, Identify and predict the error for the parameters at the current sampling time.

[0067] Based on the identified parameter vector, the estimated values ​​of the moment of inertia and the coefficient of viscous friction are obtained:

[0068]

[0069] in, For the first k The estimated value of the moment of inertia at each sampling time. For the first k The estimated value of the viscous friction coefficient at each sampling time. , They are parameter vectors The first and second components.

[0070] Through the above online parameter identification process, estimated values ​​of rotational inertia and viscous friction coefficient that match the current operating state can be obtained. Further, these estimated values ​​are output to a disturbance observer to update the mechanical motion equation parameters used by the disturbance observer, enabling the observer to estimate load disturbances based on the updated mechanical motion equations.

[0071] S104, based on the updated mechanical motion equations, the actual rotational speed and the actual q-axis current, the load disturbance of the permanent magnet synchronous motor is estimated by the disturbance observer to obtain the load disturbance estimate; a disturbance compensation current is generated according to the load disturbance estimate, and the load disturbance estimate is used as an equivalent disturbance term and introduced into the prediction model established based on the mathematical model of the permanent magnet synchronous motor.

[0072] In one implementation, the disturbance observer obtains the original load disturbance estimate based on the updated mechanical motion equations, the estimated moment of inertia, the estimated viscous friction coefficient, the actual rotational speed at the current sampling time and the previous sampling time, and the actual q-axis current at the previous sampling time.

[0073] Specifically, the estimated moment of inertia output by the online parameter identification module is used. and estimated value of viscous friction coefficient Update the parameters of the mechanical motion equations used by the disturbance observer, and construct the original load disturbance estimate based on the updated mechanical motion equations:

[0074] in, For the first k The original load disturbance estimate at each sampling time. This represents the actual q-axis current at the previous sampling time. This is the estimated value of the moment of inertia at the previous sampling time. This is the estimated value of the viscous friction coefficient at the previous sampling time. This represents the actual rotational speed at the previous sampling time. The sampling period.

[0075] Since load disturbance estimation may involve velocity differential or equivalent differential calculations, to reduce the impact of noise amplification, the original load disturbance estimate can be low-pass filtered to obtain a smoothed load disturbance estimate.

[0076] in, This is the filtered load disturbance estimate. For the first k The original load disturbance estimate at each sampling time. These are the DOB low-pass filter coefficients, and 0 < ≤1.

[0077] In one implementation, the low-pass filter coefficients can be determined based on the time constant of the disturbance observer:

[0078] in, is the time constant of the perturbation observer.

[0079] To prevent the load disturbance estimate from becoming excessively large, a saturation limit can be applied to the smoothed load disturbance estimate to obtain the final load disturbance estimate:

[0080] in, For the first k The final load disturbance estimate at each sampling time. For the first k The filtered load disturbance estimate at each sampling time point This is a saturation limiting function used to limit the input quantity within the parentheses to a preset load disturbance estimation range.

[0081] Based on the estimated load disturbance and the electromagnetic torque relationship of the permanent magnet synchronous motor, a disturbance compensation current is generated:

[0082] in, For the first k Disturbance compensation current at each sampling time Forward compensation coefficient, This is the torque coefficient.

[0083] The disturbance compensation current is subsequently combined with the base q-axis reference current and the optimal prediction correction to participate in the synthesis of the final q-axis reference current.

[0084] A disturbance compensation current is generated based on the estimated load disturbance value, and the estimated load disturbance value is used as an equivalent disturbance term and introduced into the prediction model established based on the mathematical model of the permanent magnet synchronous motor.

[0085] Please see Figure 2 The online parameter identification module outputs estimated values ​​of moment of inertia and viscous friction coefficient based on the rotational speed data and q-axis current data. These estimated values ​​are then sent to the disturbance observer to update the mechanical motion equation parameters used by the observer. The disturbance observer outputs estimated load disturbance values ​​based on the updated mechanical motion equations, actual rotational speed, and actual q-axis current.

[0086] In this embodiment, the load disturbance estimate has two functions: on the one hand, the load disturbance estimate is used to generate a disturbance compensation current, which participates in the synthesis of the final q-axis reference current; on the other hand, the load disturbance estimate is introduced as an equivalent disturbance term into the prediction model established based on the mathematical model of the permanent magnet synchronous motor, so that the prediction model considers the impact of load disturbance on future speed changes during the evaluation of candidate prediction corrections.

[0087] Specifically, an equivalent disturbance term is set in the prediction model, and the load disturbance estimate output by the disturbance observer is assigned to the equivalent disturbance term. During subsequent finite-step forward rolling prediction, the future rotational speed and future q-axis current corresponding to each candidate prediction correction are predicted based on the prediction model after introducing the equivalent disturbance term.

[0088] It should be noted that the load disturbance estimate is not a simple correction to the prediction result after the rolling prediction is completed, but rather participates in the rolling prediction process as an equivalent disturbance term in the prediction model. Therefore, when evaluating the correction amounts for each candidate prediction, the impact of load disturbances on future speed changes can be considered in advance, thereby improving the accuracy of the evaluation of candidate prediction correction amounts.

[0089] Through the above process, the online parameter identification module provides the disturbance observer with estimates of the moment of inertia and the coefficient of viscous friction. The disturbance observer provides disturbance compensation current to the current synthesis process on the one hand, and equivalent disturbance terms to the prediction model on the other hand, thereby achieving synergy between online parameter identification, disturbance observation compensation and rolling prediction optimization.

[0090] S105, for each candidate prediction correction in the finite candidate prediction correction set, construct a test q-axis reference current based on the basic q-axis reference current and the corresponding candidate prediction correction, and use the actual rotational speed and the actual q-axis current as the initial prediction values. Based on the prediction model after introducing the equivalent disturbance term, perform finite-step forward rolling prediction to obtain the future rotational speed and future q-axis current in the prediction time domain.

[0091] It should be noted that the load disturbance estimate output by the disturbance observer adopts a dual-channel approach in this application: on the one hand, the load disturbance estimate is directly introduced into the prediction model as an equivalent disturbance term, enabling the evaluation process of candidate prediction corrections to consider the impact of load disturbances on future speed changes; on the other hand, the load disturbance estimate is converted into a disturbance compensation current through the relationship between electromagnetic torque and q-axis current, and then fused with the basic q-axis reference current and the optimal prediction correction in the final q-axis reference current synthesis stage. Thus, the test q-axis reference current is used to evaluate candidate prediction corrections, the equivalent disturbance term is used to correct the prediction model, and the disturbance compensation current is used for final feedforward compensation. These three elements work together to achieve coordinated control between disturbance observation, prediction correction, and current synthesis.

[0092] Please see Figure 3 The prediction model at the current sampling time k The system receives the base q-axis reference current, a finite set of candidate prediction corrections, the current actual rotational speed, the current actual q-axis current, and the load disturbance estimate output by the disturbance observer. The load disturbance estimate serves as the equivalent disturbance term in the prediction model and participates in the finite-step forward roll prediction.

[0093] For the m-th candidate prediction correction in the finite set of candidate prediction corrections, the prediction model adds the base q-axis reference current to the candidate prediction correction to obtain the test q-axis reference current corresponding to the candidate prediction correction. For candidate prediction corrections that satisfy both current and voltage constraints, the candidate prediction correction and its dynamic weight cost function value are added to the feasible candidate set.

[0094] In one implementation, the actual rotational speed and actual q-axis current at the current sampling moment are used as initial prediction values. Specifically, the current actual rotational speed is used as the initial prediction value for rotational speed, and the current actual q-axis current is used as the initial prediction value for q-axis current. Based on the tested q-axis reference current and the prediction model after introducing an equivalent perturbation term, a finite-step forward rolling prediction is performed on the future q-axis current and future rotational speed in the prediction time domain.

[0095] The q-axis current prediction process can be determined based on the dynamic characteristics of the current loop or the q-axis current response model. Use the current actual state as the initial value for prediction:

[0096]

[0097] For h=1,2, , The q-axis current prediction model is as follows:

[0098] in, The current sampling time k Next prediction h The future q-axis current of the step, The current sampling time k Next prediction h -1 step future q-axis current, Let be the test q-axis reference current corresponding to the m-th candidate prediction correction. The equivalent current loop time constant is set to 0.0005s, or 0.5ms. For the sampling period, within the prediction time domain, the first... m The test q-axis reference current corresponding to each candidate prediction correction. It remains unchanged.

[0099] The speed prediction process can be determined based on the mechanical motion equations and equivalent disturbance terms of the permanent magnet synchronous motor:

[0100] The current sampling time k Next prediction hThe future rotation speed of the step, The current sampling time k Next prediction h -1 step future rotation speed, This is the estimated value of the moment of inertia at the current sampling time. This is the estimated value of the viscous friction coefficient at the current sampling time. The current sampling time k Next prediction h The equivalent perturbation term introduced during the step, h This indicates the prediction step number.

[0101] When no disturbance observer is introduced, take When a disturbance observer is introduced, take .in, This represents the final load disturbance estimate output by the disturbance observer at the current sampling time. Therefore, the DOB estimate is incorporated into the prediction model, enabling the model to consider the impact of load disturbances on future speed changes during the candidate prediction correction evaluation phase.

[0102] Within the prediction time domain, the prediction model can progressively calculate the future q-axis current and future rotational speed at each prediction step, obtaining the prediction result corresponding to the m-th candidate prediction correction. The prediction result includes the future rotational speed and future q-axis current within the prediction time domain. The future rotational speed is used to subsequently calculate the rotational speed prediction error and the dynamic weight cost function, while the future q-axis current is used to subsequently determine whether the current constraint is satisfied.

[0103] For each candidate prediction correction in the finite set of candidate prediction corrections, the above finite-step forward rolling prediction process is performed to obtain the prediction result corresponding to each candidate prediction correction.

[0104] S106. Based on the future speed in the prediction result and the reference speed of the corresponding prediction step, determine the speed prediction error, and dynamically adjust the weight of the cost function according to the variance of the speed prediction error to obtain the dynamic weight cost function.

[0105] The weights of the cost function are dynamically adjusted based on the variance of the speed prediction error.

[0106] For any candidate prediction correction, the prediction result is used to determine the speed prediction error based on the future speed in the prediction result and the reference speed of the corresponding prediction step. The speed prediction error characterizes the deviation of the future speed from the reference speed within the prediction time domain.

[0107] If the reference speed remains at the current smoothed reference speed within the prediction time domain, the following can be adopted: This serves as the reference speed for each prediction step. In each control cycle... k Calculate the prediction time domain Estimated variance of rotational speed prediction error within:

[0108] in, For the first k The estimated variance of the speed prediction error at each sampling time. The value is 8 to represent the length of the predicted time domain. The current sampling time k Next prediction h The future rotation speed of the step, This is the smoothed reference speed after filtering.

[0109] To avoid abrupt changes in the estimated variance of the speed prediction error due to instantaneous disturbances or noise, the estimated variance of the speed prediction error can be filtered to obtain a smoothed estimated variance.

[0110] in, For the first k The estimated variance of the smoothed prediction error at each sampling time. This is the estimated variance of the smoothed prediction error at the previous sampling time. These are the prediction error filter coefficients, and 0 ≤ ≤1, For the first k The estimated variance of the rotational speed prediction error at each sampling time.

[0111] Based on the estimated variance of the smoothed prediction error, update the weight ratio between the speed tracking error weight and the control increment penalty weight:

[0112] in, The weight ratio at the k-th sampling time. This is the weight for the speed tracking error, with an initial value of 1.6. To control the incremental penalty weight.

[0113] Update the weight ratio based on the estimated variance of the smoothed prediction error:

[0114] in, This is the steady-state ratio, with a value of 2.0. For adaptive gain, the value is 150. As a reference for the prediction error variance threshold, and >0, specifically a value of 25 (r / min). 2 .

[0115] To avoid the weight ratio being too large or too small, the weight ratio can be limited:

[0116] in, and These are the lower and upper limits of the weight ratio, respectively, and 0 < ≤ The specific value is r max =10, r min =0.5.

[0117] In one implementation, the control increment penalty weight is kept at the baseline value:

[0118] in, The preset baseline value for controlling the incremental penalty weight is set to 0.8.

[0119] Update the speed tracking error weights based on the weight ratios:

[0120] When the estimated variance of the smoothed prediction error increases, the weight ratio... As the speed increases, the prediction model pays more attention to the speed tracking error; when the estimated variance of the smoothed prediction error is small, the weight ratio is kept at a low level to avoid excessive control increment.

[0121] Based on the weighted ratio after limiting, the speed tracking error weight and the control increment penalty weight are determined, and combined with the q-axis current penalty weight, a dynamic weight cost function is constructed:

[0122] in, For the first m The cost function value corresponding to each candidate prediction correction. This is the q-axis current penalty weight, with a value of 0.1. For the first h The speed prediction error at each prediction step differs from the speed error at the current sampling time. , For the first m Candidate prediction correction The current sampling time k Next prediction h The future q-axis current of the step, and:

[0123] In the formula, the dynamic weight cost function includes a speed tracking error term, a control increment penalty term, and a q-axis current penalty term.

[0124] S107 determines the optimal prediction correction amount based on current constraints, voltage constraints, and dynamic weight cost function.

[0125] For each candidate prediction correction, the future q-axis current in the prediction time domain is first determined to be within the preset q-axis current constraint range. The current constraint can be expressed as: , h =1,2, ,

[0126] in, The current sampling time k Next prediction h The future q-axis current of the step, , These are the lower and upper limits of the q-axis current, respectively. If the future q-axis current in any prediction step exceeds the q-axis current constraint range, then the candidate prediction correction is determined to not meet the current constraint.

[0127] Furthermore, based on the test q-axis reference current corresponding to the candidate predicted correction, the future q-axis current in the prediction time domain, and the mathematical model of the permanent magnet synchronous motor, the predicted voltage in the prediction time domain is determined, and it is judged whether the predicted voltage meets the inverter's allowable output voltage amplitude constraint. The voltage constraint can be expressed as:

[0128] in, , These are the predicted d-axis voltage and predicted q-axis voltage, calculated in the prediction time domain based on the mathematical model of the permanent magnet synchronous motor, respectively. This refers to the allowable output voltage amplitude of the inverter.

[0129] If the future q-axis current corresponding to a candidate prediction correction does not meet the current constraint, or the corresponding prediction voltage does not meet the voltage constraint, then the candidate prediction correction is determined to be an infeasible solution and will not participate in the subsequent optimal candidate comparison.

[0130] For candidate prediction corrections that satisfy both current and voltage constraints, their corresponding dynamic weight cost function values ​​are calculated. Then, the dynamic weight cost function values ​​corresponding to each feasible candidate prediction correction are compared, and the candidate prediction correction with the smallest dynamic weight cost function value is determined as the optimal prediction correction.

[0131] The optimal prediction correction amount can be expressed as:

[0132] in, The current sampling time k The determined optimal prediction correction amount For the first m Each candidate prediction correction The current sampling time k A finite set of candidate prediction corrections For the first m The dynamic weight cost function value corresponding to each candidate prediction correction. This indicates that the candidate prediction correction amount that minimizes the cost function value is selected from the set of finite candidate prediction correction amounts.

[0133] Therefore, the prediction model determines the optimal prediction correction amount through rolling prediction of a finite candidate set and constraint optimization.

[0134] S108 integrates the basic q-axis reference current, the optimal prediction correction amount, and the disturbance compensation current to generate the final q-axis reference current, and uses the final q-axis reference current as the q-axis current command of the current inner loop to control the speed of the permanent magnet synchronous motor.

[0135] After obtaining the basic q-axis reference current, the optimal prediction correction, and the disturbance compensation current, the three are fused to obtain the final q-axis reference current.

[0136] Among them, the basic q-axis reference current is obtained from the basic q-axis reference current generation process and is used to provide the reference control quantity for speed closed-loop control; the optimal prediction correction quantity is determined by the prediction model from a finite set of candidate prediction correction quantities through rolling prediction, constraint judgment and dynamic weight cost function evaluation, and is used to predict, optimize and correct the basic q-axis reference current; the disturbance compensation current is generated by the disturbance observer based on the load disturbance estimate and is used to compensate for the impact of load disturbance on the speed control process.

[0137] In one implementation, the base q-axis reference current, the optimal prediction correction, and the disturbance compensation current can be superimposed, and the superposition result can be limited to obtain the final q-axis reference current.

[0138] in:

[0139] This is the final q-axis reference current. Based on the q-axis reference current, This is the optimal prediction correction amount. For disturbance compensation current, It is a saturation limiting function. Forward compensation coefficient, This represents the final load disturbance estimate at the current sampling time k. This is the torque coefficient.

[0140] Through the above fusion process, the final q-axis reference current simultaneously includes the speed error feedback control quantity, the finite candidate set rolling optimization correction quantity, and the load disturbance compensation quantity. Compared with the control method that relies solely on the output of the q-axis reference current by the speed PI controller, the embodiments of this application can more promptly correct the q-axis reference current under the influence of changes in reference speed or load disturbances, thereby improving the speed tracking performance and disturbance suppression capability of the permanent magnet synchronous motor.

[0141] After obtaining the final q-axis reference current, it is used as the q-axis current command for the inner current loop to control the speed of the permanent magnet synchronous motor (PMSM). In one implementation, the inner current loop can generate a voltage control quantity based on the final q-axis reference current and the actual q-axis current, and apply it to the PMSM through an inverter or pulse width modulation module to adjust the electromagnetic torque and actual speed of the PMSM. The actual speed and actual q-axis current during PMSM operation are used as feedback data for further generation of the basic q-axis reference current, finite-step forward roll prediction, online parameter identification, and disturbance observation and compensation.

[0142] Example 2 Please see Figure 4 This application also provides a structure for a closed-loop speed control system for a permanent magnet synchronous motor. Figure 4 This is used to illustrate the signal flow relationship of the speed control method provided in the embodiments of this application in the closed-loop control system of a permanent magnet synchronous motor.

[0143] like Figure 4 As shown, the speed closed-loop control system may include a speed controller, a current inner loop, an inverter or pulse width modulation module, and a permanent magnet synchronous motor. The speed controller generates a final q-axis reference current based on a reference speed and feedback data. The current inner loop uses the final q-axis reference current as a q-axis current command and combines it with the actual q-axis current for current closed-loop control. The inverter or pulse width modulation module drives the permanent magnet synchronous motor based on the voltage control output from the current inner loop. The actual speed and actual q-axis current generated during the operation of the permanent magnet synchronous motor are returned to the speed controller as feedback data.

[0144] In one embodiment, the speed controller includes a basic q-axis reference current generation unit, a model prediction control unit, an online parameter identification unit, a disturbance observation and compensation unit, and a current synthesis unit.

[0145] The basic q-axis reference current generation unit receives the reference rotational speed and the actual rotational speed from the feedback data, and generates a basic q-axis reference current based on the speed error between the reference rotational speed and the actual rotational speed. This basic q-axis reference current is input to the current synthesis unit as the reference component of the final q-axis reference current; it is also input to the model prediction control unit to construct the test q-axis reference current corresponding to each candidate prediction correction.

[0146] The model prediction control unit receives the basic q-axis reference current, a finite set of candidate prediction corrections, feedback data, and the equivalent disturbance term output by the disturbance observation and compensation unit. The feedback data may include the actual rotational speed and actual q-axis current, and serves as the initial prediction value in the rolling prediction process. The model prediction control unit constructs a test q-axis reference current for each candidate prediction correction, and performs a finite-step forward rolling prediction based on the prediction model after introducing the equivalent disturbance term, obtaining the future rotational speed and future q-axis current in the prediction time domain. The model prediction control unit also determines the current constraint based on the future q-axis current, determines the prediction voltage and determines the voltage constraint based on the test q-axis reference current, the future q-axis current, and the permanent magnet synchronous motor mathematical model, and determines the optimal prediction correction amount by combining the dynamic weight cost function.

[0147] The online parameter identification unit receives the rotational speed data and q-axis current data from the feedback data, and identifies the estimated values ​​of the moment of inertia and viscous friction coefficient online based on the discrete form of the mechanical motion equations of the permanent magnet synchronous motor. The estimated values ​​of the moment of inertia and viscous friction coefficient are output to the disturbance observation and compensation unit to update the mechanical motion equation parameters used by the disturbance observation and compensation unit.

[0148] The disturbance observation and compensation unit receives the estimated moment of inertia and viscous friction coefficient from the online parameter identification unit, and also receives the actual rotational speed and actual q-axis current from the feedback data. Based on the updated mechanical motion equations, actual rotational speed, and actual q-axis current, the disturbance observation and compensation unit estimates the load disturbance of the permanent magnet synchronous motor, obtaining a load disturbance estimate. This load disturbance estimate is used, on the one hand, to generate a disturbance compensation current, which is output to the current synthesis unit; on the other hand, it is output as an equivalent disturbance term to the prediction model in the model prediction and control unit.

[0149] The current synthesis unit receives the base q-axis reference current, the optimal prediction correction, and the disturbance compensation current, and fuses these three to generate the final q-axis reference current. This final q-axis reference current serves as the q-axis current command for the inner current loop, used for speed control of the permanent magnet synchronous motor.

[0150] The actual speed and actual q-axis current generated during the operation of the permanent magnet synchronous motor are returned to the speed controller as feedback data. This feedback data can be input to the basic q-axis reference current generation unit, the model prediction control unit, the online parameter identification unit, and the disturbance observation and compensation unit, respectively, for speed error calculation, rolling prediction initial value setting, online parameter identification, and load disturbance estimation.

[0151] pass Figure 4 The system structure shown in this application integrates basic speed feedback control, finite candidate set rolling optimization, online parameter identification, and disturbance observation and compensation into the same speed closed-loop control process. Specifically, the basic q-axis reference current generation unit provides the reference control quantity, the model prediction control unit provides the optimal prediction correction quantity, the disturbance observation and compensation unit provides the disturbance compensation current, and the current synthesis unit fuses these three components to form the final q-axis reference current, thereby improving the dynamic response and disturbance rejection capabilities of the permanent magnet synchronous motor speed control system.

[0152] Example 3 To verify the control effect of the permanent magnet synchronous motor speed control method provided in this application embodiment, in one implementation, the surface-mounted PMSM speed control system can be simulated and verified in the MATLAB simulation environment. The simulation verification is used to illustrate the speed tracking performance, torque response capability, disturbance estimation effect, and disturbance compensation effect of the method under reference speed variation and load disturbance conditions, and is not intended to limit the scope of protection of this application.

[0153] In this embodiment, the simulation object is a surface-mount permanent magnet synchronous motor, and the simulation uses a set of typical surface-mount PMSM parameters, the main parameters of which are shown in Table 1.

[0154]

[0155] In this embodiment, the total simulation time is set to 0.30 s. The reference speed jumps from 0 to 1000 r / min at t=0.005 s, and a load torque disturbance of 1 N·m is applied at t=0.20 s to simulate the dynamic response process of the permanent magnet synchronous motor speed control system under start-up tracking and sudden load conditions. The above parameters and operating conditions are only used to illustrate the control effect of the embodiments of this application and are not intended to limit the scope of protection of this application.

[0156] This embodiment sets up three control strategies for comparison: The first control strategy is proportional-integral control, or PI control. PI stands for Proportional Integral. Under PI control, the speed loop primarily generates the q-axis reference current based on the speed error between the reference speed and the actual speed. The second control strategy is MPC+DOB control. MPC stands for Model Predictive Control, and DOB stands for Disturbance Observer. The MPC+DOB control strategy introduces finite prediction and disturbance observation compensation on top of PI control. The third control strategy, "Proposed," refers to the permanent magnet synchronous motor speed control method provided in this application.

[0157] Please see Figure 5 , Figure 5 The speed response curves are shown under different control strategies. Figure 5 In the graph, the horizontal axis represents time in seconds (s); the vertical axis represents the speed of the permanent magnet synchronous motor in r / min; and "reference value" indicates the reference speed.

[0158] Please see Figure 6 , Figure 6 The electromagnetic torque response curves are shown under different control strategies. The horizontal axis represents time in seconds; the vertical axis represents the electromagnetic torque of the permanent magnet synchronous motor in N·m; "Load Torque" represents the load torque disturbance applied during the simulation.

[0159] Please see Figure 7 , Figure 7 This is the curve for estimating load disturbance. Figure 7 In the graph, the horizontal axis represents time in seconds (s), and the vertical axis represents the estimated load disturbance in N·m. Figure 7 The dashed lines in the graph represent the actual load disturbance, and the solid lines represent the load disturbance estimates output by the disturbance observer. If a DOB appears in the graph, it represents the disturbance observer. Figure 7 It can be seen that the disturbance observer can estimate the load disturbance, and the estimated load disturbance can follow the actual load disturbance.

[0160] Please see Figure 8 , Figure 8 This is the q-axis disturbance compensation current curve. Figure 8 In the graph, the horizontal axis represents time in seconds (s), and the vertical axis represents the q-axis disturbance compensation current in amperes (A). The q-axis disturbance compensation current is the compensation amount generated by the disturbance observer based on the estimated load disturbance. Figure 8 It can be seen that after the load disturbance is applied, the q-axis disturbance compensation current is established as the load disturbance estimate changes, and is used to compensate for the influence of the load disturbance on the speed control process.

[0161] In conclusion, Figures 5 to 8 The control effects of the permanent magnet synchronous motor speed control method provided in this application are explained from four aspects: speed response, electromagnetic torque response, load disturbance estimation, and q-axis disturbance compensation current. Among them, Figure 5 and Figure 6 To illustrate that, compared to PI control, the method of this application can improve the speed recovery capability under load disturbances by utilizing rolling prediction, disturbance observation compensation, and online parameter correction; compared to MPC+DOB control, the method of this application further utilizes adaptive weight adjustment and online parameter identification, so that the load disturbance estimate output by the disturbance observer simultaneously participates in the generation of compensation current and the correction of prediction model, thereby improving the accuracy of candidate prediction correction evaluation and the stability of speed closed-loop control. Figure 7 and Figure 8 This is used to illustrate that the disturbance observer can estimate load disturbances and generate a q-axis disturbance compensation current based on the estimated load disturbance. The above simulation results are only used to illustrate the control effect of the method in this application and are not intended to limit the scope of protection of this application.

[0162] Example 4 Please see Figure 9 This application also provides a speed control device for a permanent magnet synchronous motor. The speed control device can be used to execute the speed control method for a permanent magnet synchronous motor in the foregoing embodiments. The device includes a model acquisition module, a basic current and candidate quantity generation module, an online parameter identification module, a disturbance observation and compensation module, a rolling prediction module, an adaptive weight adjustment module, a constraint optimization module, and a current synthesis control module.

[0163] The model acquisition module is used to acquire the mathematical model of the permanent magnet synchronous motor, as well as the reference speed, actual speed, and actual q-axis current at the current sampling time; The basic current and candidate quantity generation module is used to generate a basic q-axis reference current based on the speed error between the reference speed and the actual speed, and to construct a finite set of candidate prediction correction quantities. The online parameter identification module is used to identify the estimated values ​​of rotational inertia and viscous friction coefficient online based on the rotational speed data and q-axis current data during the operation of the permanent magnet synchronous motor, and update the mechanical motion equation parameters adopted by the disturbance observer based on the estimated values ​​of rotational inertia and viscous friction coefficient. The disturbance observation and compensation module is used to estimate the load disturbance of the permanent magnet synchronous motor based on the updated mechanical motion equations, the actual rotational speed, and the actual q-axis current through the disturbance observer, and obtain the load disturbance estimate; generate a disturbance compensation current based on the load disturbance estimate, and introduce the load disturbance estimate as an equivalent disturbance term into the prediction model established based on the mathematical model of the permanent magnet synchronous motor; The rolling prediction module is used to construct a test q-axis reference current based on the base q-axis reference current and the corresponding candidate prediction correction for each candidate prediction correction in the finite candidate prediction correction set, and to perform finite-step forward rolling prediction based on the prediction model after introducing the equivalent disturbance term, using the actual rotation speed and the actual q-axis current as the prediction initial values, to obtain the future rotation speed and future q-axis current in the prediction time domain. An adaptive weight adjustment module is used to determine the speed prediction error based on the future speed and the reference speed of the corresponding prediction step, and dynamically adjust the weight of the cost function based on the variance of the speed prediction error to obtain a dynamic weight cost function. The constraint optimization module is used to determine the current constraint based on the future q-axis current, determine the predicted voltage in the prediction time domain based on the test q-axis reference current, the future q-axis current, and the mathematical model of the permanent magnet synchronous motor, and determine the voltage constraint based on the predicted voltage; and determine the optimal prediction correction amount from the finite candidate prediction correction amount set based on the current constraint, the voltage constraint, and the dynamic weight cost function. The current synthesis control module is used to fuse the basic q-axis reference current, the optimal prediction correction amount, and the disturbance compensation current to generate the final q-axis reference current, and use the final q-axis reference current as the q-axis current command of the current inner loop to control the speed of the permanent magnet synchronous motor.

[0164] exist Figure 9 In the device structure shown, the model acquisition module provides corresponding model data and sampling data to the basic current and candidate quantity generation module, the rolling prediction module, and the online parameter identification module; the basic current and candidate quantity generation module provides the basic q-axis reference current to the rolling prediction module and the current synthesis control module, and provides a finite set of candidate prediction corrections to the rolling prediction module; the online parameter identification module provides the estimated values ​​of rotational inertia and viscous friction coefficient to the disturbance observation and compensation module; the disturbance observation and compensation module provides the equivalent disturbance term to the rolling prediction module, and provides the disturbance compensation current to the current synthesis control module; the rolling prediction module provides the prediction results to the adaptive weight adjustment module and the constraint optimization module; the adaptive weight adjustment module provides the dynamic weight cost function to the constraint optimization module; the constraint optimization module provides the optimal prediction correction to the current synthesis control module; and the current synthesis control module outputs the final q-axis reference current.

[0165] It should be noted that the permanent magnet synchronous motor speed control device in this embodiment can be implemented in hardware, as a software functional module, or a combination of hardware and software. The modules can be integrated into the same controller or distributed according to the actual control system structure. The specific functions and implementation processes of each module can be found in the relevant descriptions in the foregoing method embodiments, and will not be repeated here.

[0166] The above description is merely a specific embodiment of this application and is not intended to limit the scope of protection of this application. Any equivalent substitutions, modifications, or variations made to the embodiments of this application within the technical concept and principles of this application should be included within the scope of protection of this application.

Claims

1. A speed control method for a permanent magnet synchronous motor, characterized in that, include: Obtain the mathematical model of the permanent magnet synchronous motor, as well as the reference speed, actual speed, and actual q-axis current at the current sampling moment; Based on the speed error between the reference rotational speed and the actual rotational speed, a basic q-axis reference current is generated, and a finite set of candidate prediction corrections is constructed. Based on the speed data and q-axis current data during the operation of the permanent magnet synchronous motor, the estimated values ​​of the moment of inertia and the viscous friction coefficient are identified online, and the mechanical motion equation parameters used by the disturbance observer are updated based on the estimated values ​​of the moment of inertia and the viscous friction coefficient. Based on the updated mechanical motion equations, the actual rotational speed, and the actual q-axis current, the load disturbance of the permanent magnet synchronous motor is estimated using the disturbance observer to obtain a load disturbance estimate. A disturbance compensation current is generated based on the load disturbance estimate, and the load disturbance estimate is used as an equivalent disturbance term and introduced into the prediction model established based on the mathematical model of the permanent magnet synchronous motor. For each candidate prediction correction in the finite candidate prediction correction set, a test q-axis reference current is constructed based on the basic q-axis reference current and the corresponding candidate prediction correction. The actual rotational speed and the actual q-axis current are used as the initial prediction values. Based on the prediction model after introducing the equivalent disturbance term, a finite-step forward rolling prediction is performed to obtain the future rotational speed and future q-axis current in the prediction time domain. The speed prediction error is determined based on the future speed and the reference speed of the corresponding prediction step, and the weight of the cost function is dynamically adjusted based on the variance of the speed prediction error to obtain the dynamic weight cost function. The current constraint is determined based on the future q-axis current; the predicted voltage in the prediction time domain is determined based on the test q-axis reference current, the future q-axis current, and the mathematical model of the permanent magnet synchronous motor; and the voltage constraint is determined based on the predicted voltage; based on the current constraint, the voltage constraint, and the dynamic weight cost function, the optimal prediction correction amount is determined from the set of finite candidate prediction correction amounts. The base q-axis reference current, the optimal prediction correction, and the disturbance compensation current are combined to generate the final q-axis reference current, which is then used as the q-axis current command for the inner current loop to control the speed of the permanent magnet synchronous motor.

2. The speed control method for a permanent magnet synchronous motor according to claim 1, characterized in that, The mathematical model of the permanent magnet synchronous motor includes the voltage equation, electromagnetic torque equation, and mechanical motion equation in the synchronous rotating dq coordinate system. The voltage equation is used to characterize the relationship between d-axis voltage, q-axis voltage, d-axis current, q-axis current, inductive component, stator resistance, permanent magnet flux linkage, and electric angular velocity. The electromagnetic torque equation is used to characterize the relationship between the q-axis current and the electromagnetic torque. The mechanical motion equations are used to characterize the relationship between electromagnetic torque, moment of inertia, mechanical angular velocity, viscous friction coefficient and load disturbance, and are used in the prediction model for velocity prediction, the disturbance observer for load disturbance estimation, and online identification of estimated moment of inertia and viscous friction coefficient values.

3. The speed control method for a permanent magnet synchronous motor according to claim 1, characterized in that, The step of generating a basic q-axis reference current based on the speed error between the reference rotational speed and the actual rotational speed, and constructing a finite set of candidate prediction corrections, includes: A first-order filter is applied to the original reference speed to obtain a smoothed reference speed; The speed error is obtained based on the difference between the smoothed reference speed and the actual speed; The speed error is input into the speed proportional-integral controller to obtain the unlimited base q-axis reference current; The unlimited base q-axis reference current is limited to obtain the base q-axis reference current; At the current sampling time, multiple candidate prediction corrections are generated based on a preset search interval and a preset search step size to form the finite set of candidate prediction corrections.

4. The speed control method for a permanent magnet synchronous motor according to claim 1, characterized in that, The process of identifying estimated values ​​for the moment of inertia and the coefficient of viscous friction online based on the speed data and q-axis current data during the operation of the permanent magnet synchronous motor includes: Constructing a parameter identification regression model based on the discrete form of the mechanical motion equation; The parameter vector in the parameter identification regression model is updated online using a recursive least squares algorithm with a forgetting factor. The estimated value of the moment of inertia and the estimated value of the viscous friction coefficient are determined based on the updated parameter vector; The estimated values ​​of rotational inertia and viscous friction coefficient are subject to upper and lower limits or rate of change limits, and the limited estimated values ​​of rotational inertia and viscous friction coefficient are output to the disturbance observer to update the mechanical motion equation parameters used by the disturbance observer.

5. The speed control method for a permanent magnet synchronous motor according to claim 1, characterized in that, Based on the updated mechanical motion equations, the actual rotational speed, and the actual q-axis current, the load disturbance of the permanent magnet synchronous motor is estimated using the disturbance observer to obtain the load disturbance estimate. Generating a disturbance compensation current based on the estimated load disturbance includes: The disturbance observer, based on the updated mechanical motion equations, obtains the original load disturbance estimate according to the estimated moment of inertia, the estimated viscous friction coefficient, the mechanical angular velocity corresponding to the current actual rotational speed, and the actual q-axis current. The original load disturbance estimate is low-pass filtered to obtain a smoothed load disturbance estimate; The smoothed load disturbance estimate is subjected to saturation constraint to obtain the load disturbance estimate; The disturbance compensation current is generated based on the estimated load disturbance and the feedforward compensation coefficient.

6. The speed control method for a permanent magnet synchronous motor according to claim 1, characterized in that, The step of incorporating the estimated load disturbance as an equivalent disturbance term into the prediction model established based on the mathematical model of the permanent magnet synchronous motor includes: An equivalent perturbation term is set in the prediction model; The load disturbance estimate output by the disturbance observer is assigned to the equivalent disturbance term; Based on the prediction model after introducing the equivalent disturbance term, a finite-step forward rolling prediction is performed on the future speed corresponding to each candidate prediction correction, so that the prediction model considers the impact of load disturbance on future speed changes during the evaluation of candidate prediction corrections.

7. The speed control method for a permanent magnet synchronous motor according to claim 1, characterized in that, For each candidate prediction correction in the finite candidate prediction correction set, a test q-axis reference current is constructed based on the base q-axis reference current and the corresponding candidate prediction correction. Using the actual rotational speed and the actual q-axis current as initial prediction values, a finite-step forward rolling prediction is performed based on the prediction model after introducing the equivalent disturbance term to obtain the future rotational speed and future q-axis current in the prediction time domain, including: The basic q-axis reference current is added to any candidate prediction correction in the set of finite candidate prediction corrections to obtain the test q-axis reference current corresponding to the candidate prediction correction. Using the actual rotational speed and actual q-axis current at the current sampling moment as the initial prediction values, and based on the test q-axis reference current and the prediction model after introducing the equivalent disturbance term, a finite-step forward rolling prediction is performed on the future q-axis current and future rotational speed in the prediction time domain to obtain the future rotational speed and future q-axis current in the prediction time domain corresponding to the candidate prediction correction amount.

8. The speed control method for a permanent magnet synchronous motor according to claim 1, characterized in that, The step of dynamically adjusting the weights of the cost function based on the variance of the speed prediction error to obtain the dynamic weight cost function includes: Calculate the variance of the speed prediction error in the prediction time domain; The variance of the speed prediction error is filtered to obtain a smoothed estimate of the prediction error variance. Update the weight ratio between the speed tracking error weight and the control increment penalty weight based on the smoothed prediction error variance estimate; The weight ratio is limited, and the speed tracking error weight and the control increment penalty weight are determined based on the limited weight ratio. The dynamic weight cost function is constructed based on the speed tracking error weight, the control increment penalty weight, and the q-axis current penalty weight.

9. The speed control method for a permanent magnet synchronous motor according to claim 1, characterized in that, The step of determining the optimal prediction correction amount from the finite candidate prediction correction amount set based on current constraints, voltage constraints, and the dynamic weight cost function includes: For each candidate prediction correction, determine whether the future q-axis current in the prediction time domain is within the preset q-axis current constraint range; and determine the prediction voltage in the prediction time domain based on the test q-axis reference current, the future q-axis current, and the mathematical model of the permanent magnet synchronous motor corresponding to the candidate prediction correction, and determine whether the prediction voltage meets the inverter's allowable output voltage amplitude constraint. Candidate prediction corrections that do not meet the current constraint or the voltage constraint are determined to be infeasible solutions; For candidate prediction corrections that satisfy the current and voltage constraints, calculate their corresponding dynamic weight cost function values. Compare the dynamic weight cost function values ​​corresponding to each feasible candidate prediction correction amount, and determine the candidate prediction correction amount with the smallest dynamic weight cost function value as the optimal prediction correction amount.

10. A speed control device for a permanent magnet synchronous motor, characterized in that, include: The model acquisition module is used to acquire the mathematical model of the permanent magnet synchronous motor, as well as the reference speed, actual speed, and actual q-axis current at the current sampling time; The basic current and candidate quantity generation module is used to generate a basic q-axis reference current based on the speed error between the reference speed and the actual speed, and to construct a finite set of candidate prediction correction quantities. The online parameter identification module is used to identify the estimated values ​​of rotational inertia and viscous friction coefficient online based on the rotational speed data and q-axis current data during the operation of the permanent magnet synchronous motor, and update the mechanical motion equation parameters adopted by the disturbance observer based on the estimated values ​​of rotational inertia and viscous friction coefficient. The disturbance observation and compensation module is used to estimate the load disturbance of the permanent magnet synchronous motor based on the updated mechanical motion equations, the actual rotational speed, and the actual q-axis current through the disturbance observer, and obtain the load disturbance estimate; generate a disturbance compensation current based on the load disturbance estimate, and introduce the load disturbance estimate as an equivalent disturbance term into the prediction model established based on the mathematical model of the permanent magnet synchronous motor; The rolling prediction module is used to construct a test q-axis reference current based on the base q-axis reference current and the corresponding candidate prediction correction for each candidate prediction correction in the finite candidate prediction correction set, and to perform finite-step forward rolling prediction based on the prediction model after introducing the equivalent disturbance term, using the actual rotation speed and the actual q-axis current as the prediction initial values, to obtain the future rotation speed and future q-axis current in the prediction time domain. An adaptive weight adjustment module is used to determine the speed prediction error based on the future speed and the reference speed of the corresponding prediction step, and dynamically adjust the weight of the cost function based on the variance of the speed prediction error to obtain a dynamic weight cost function. The constraint optimization module is used to determine the current constraint based on the future q-axis current, determine the predicted voltage in the prediction time domain based on the test q-axis reference current, the future q-axis current, and the mathematical model of the permanent magnet synchronous motor, and determine the voltage constraint based on the predicted voltage; and determine the optimal prediction correction amount from the finite candidate prediction correction amount set based on the current constraint, the voltage constraint, and the dynamic weight cost function. The current synthesis control module is used to fuse the basic q-axis reference current, the optimal prediction correction amount, and the disturbance compensation current to generate the final q-axis reference current, and use the final q-axis reference current as the q-axis current command of the current inner loop to control the speed of the permanent magnet synchronous motor.

Images