Current prediction control method for estimating inductance parameters based on observer

Abstract

The invention relates to a current prediction control method for estimating inductance parameters based on an observer, and aims to solve the problems that the traditional dead-beat current prediction control depends on accurate motor parameters, the tracking error is large when the parameters are mismatched, the gain adjustment of a Luenberger observer is difficult, a fal function is easy to jitter and oscillate, and online parameter identification is not available. The design comprises the following steps: firstly, deducing a permanent magnet synchronous motor discrete mathematical model and analyzing parameter sensitivity; then, disturbance is concentrated in the hyper-local model, and a nonlinear Fun function of an observer estimation error and disturbance gain is established to replace a fal function; designing a variable-gain Luenberger observer, and analyzing the stability through a Michilia stability criterion; and finally, on-line identification of inductance parameters is carried out by using the disturbance quantity of the observer, and the optimal reference voltage is generated. Experiments show that when the inductance is mismatched, the current tracking error is obviously reduced, the recovery is faster after the load is suddenly added, and the robustness and the dynamic response precision of the system are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of permanent magnet synchronous motor control technology, and in particular to a current prediction control method based on observer-estimated inductance parameters. Background Technology

[0002] Permanent magnet synchronous motors (PMSMs) are widely used in new energy vehicles, industrial servos, and aerospace due to their high power density, high efficiency, and excellent dynamic response characteristics. To achieve high-precision control of motor current and electromagnetic torque, model predictive control (MMC) has attracted widespread attention because it can directly consider system constraints and optimization objectives. Deadbeat predictive current control, in particular, demonstrates significant advantages in high-performance drive systems due to its one-step prediction and fast dynamic response characteristics. Deadbeat predictive current control heavily relies on accurate motor parameter models, including inductance, stator resistance, and permanent magnet flux linkage. However, in actual operation, these parameters are affected by factors such as temperature changes, magnetic saturation effects, and flux linkage degradation, causing deviations between the predictive model and the actual system. This leads to decreased control performance and may even cause system instability. Therefore, parameter uncertainty and external disturbances have become major bottlenecks limiting the further application of deadbeat predictive control.

[0003] To address this issue, researchers have primarily employed cost function construction, parameter identification, and observer construction methods to improve the accuracy of current prediction. Among these, the observer construction method is widely used due to its superior performance. This method can estimate system disturbances in real time, achieve model error compensation, and exhibit better robustness and dynamic performance under complex operating conditions. However, the observer parameters are difficult to tune, and the gain design significantly impacts the observer's performance.

[0004] Existing methods use a hyperbolic tangent function to replace observer gain to suppress initial disturbance compensation; however, the design of this function causes the disturbance gain to converge to a fixed value over time. While replacing observer gain with an arctangent function can reduce the impact of motor parameter mismatch, this method has poor anti-interference performance and does not identify motor parameters. However, during motor operation, sudden loads and accelerations can introduce significant disturbances, and parameter mismatch requires strong motor robustness. Better disturbance gain and more accurate motor parameters lead to more precise current prediction. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a current predictive control method based on observer-estimated inductor parameters. This method solves the problems in existing permanent magnet synchronous motor current predictive control, such as difficulty in adjusting the Luneburger observer gain, the tendency of the fal function to induce discrete jitter, strong parameter dependence, and the lack of online identification leading to decreased control accuracy when parameters are mismatched. This invention designs a variable-gain Luneburger observer based on a nonlinear Fun function and uses the observer disturbance to identify inductor parameters online, effectively compensating for parameter mismatch and improving system robustness and current tracking accuracy. When the motor is disturbed, this method adjusts the Luneburger observer disturbance gain online based on the observer estimation error, constructs a Luneburger observer based on the online adjusted disturbance gain for deadbeat-free current predictive control, and simultaneously estimates the inductance using the observer disturbance to obtain the optimal control voltage.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a current prediction control method based on observer-estimated inductance parameters, comprising the following steps: A discrete current prediction model of a permanent magnet synchronous motor in the dq-axis coordinate system is constructed, the voltage equation for deadbeat current prediction control is derived, and the sensitivity of the motor parameters resistance, inductance and permanent magnet flux mismatch to current tracking performance is analyzed to obtain the control voltage for parameter mismatch. The parameter mismatch and external disturbances during motor operation are collectively defined as a unified disturbance. Based on the hyperlocal model, the control voltage equation for parameter mismatch is transformed into a current dynamic equation; Design a nonlinear function relating the observer estimation error to the perturbation gain. Construct a variable-gain Romberg continuous observer based on this nonlinear function. Analyze the stability of the observer in the discrete domain using the Julie stability criterion and determine the effective range of values ​​for the gain parameter. The inductance parameters of the motor are identified in real time by using the disturbance output of the discrete domain variable gain Luneburg observer. The estimated inductance parameters obtained by identification are substituted into the initial reference voltage calculation formula to generate the optimal reference voltage to compensate for parameter mismatch, thereby realizing high-precision current predictive control of permanent magnet synchronous motor.

[0007] Furthermore, a discrete current prediction model for the permanent magnet synchronous motor in the dq-axis coordinate system is constructed. The specific process includes: Establish the dynamic equation for the stator current of the permanent magnet synchronous motor: (1) Matrix definition: , , State variables: For the dq-axis stator current, the input variable is: This refers to the dq-axis stator voltage. For the d-axis stator inductance, For the q-axis stator inductance, Let be the angular velocity of the motor. R is the actual magnetic flux linkage of the rotor permanent magnet during operation, and R is the actual resistance of the motor during operation. pass A simplified matrix is ​​obtained by simplifying the surface-mount motor, where L represents the inductance parameter of the motor during actual operation, and Euler discretization is used, based on the current... Predicting stator current information at the next time step by using stator current at a given time step transforms the continuous model into a discrete current prediction model: (2) The simplified matrix after discretization is: , , in, The dq-axis stator current at the next moment. For the present The dq-axis stator current at time t.

[0008] Furthermore, the specific calculation process for obtaining the control voltage for parameter mismatch includes: By selecting the inverter output voltage, the dq-axis stator current at the next moment is... Accurately track the reference current in the next sampling period ,Right now: (3) current Stator voltage along the dq axis at time 1 The reference current of the dq axis at the next moment. , Substituting the next moment's dq-axis stator current into formula (2), the optimal control voltage equation in the deadbeat current predictive control (DPCC) of the permanent magnet synchronous motor (PMSM) under parameter mismatch is calculated as follows: (4) in, , , The corresponding discrete matrix; Through the current The optimal control voltage required to predict the stator current at the next moment is obtained based on the surface-mounted motor unfolded into a dq-axis configuration. , As a prediction of the present Optimal control voltage is maintained at all times to control the inverter's output power and drive the motor to track the reference current. (5) in, , This is the resistance voltage drop term. , for The rate of change of current in the inductor required to track the reference current during the sampling time. , For dq axis cross-coupling terms, This is the back electromotive force term of the permanent magnet; During stable motor operation, when parameter mismatch occurs: , , The control voltage equation for the parameter mismatch is calculated as follows: (6) in, , , These are the nominal parameters of the motor's resistance, inductance, and flux linkage. , , This refers to the mismatch in resistance, inductance, and flux linkage during motor operation.

[0009] Furthermore, the parameter mismatch and external disturbances during motor operation are collectively defined as a unified disturbance quantity. Specifically, it includes: The parameter mismatches of resistance, inductance, and flux linkage during motor operation, along with external disturbances, are aggregated into a unified disturbance. To simplify the observer design, the perturbation set is defined as follows: (7) in, This represents the disturbance along the dq axis. This represents the perturbation term due to cross-coupling and flux variation; , , This refers to the mismatch in resistance, inductance, and flux linkage during motor operation. For the dq axis stator current, Let be the derivative of the dq-axis stator current. This represents the angular velocity of the motor. The control voltage equation after parameter mismatch is transformed into: (8) in, , , is the corresponding discrete matrix.

[0010] Furthermore, based on the hyperlocal model, the control voltage equation for parameter mismatch is transformed into a current dynamic equation, specifically including: Based on hyperlocal model ,in The gain of the dq-axis control voltage. Let dq-axis disturbance be represented. A direct relationship is established between the rate of change of current and the control voltage and disturbance. The rate of change of current is a linear combination of the disturbance and the control voltage. The control voltage equation (8) of the transformed parameter mismatch is converted into the current dynamic equation, specifically in the form of the current derivative: (9) in, , To control the contribution of voltage to changes in nominal inductance parameters, The combined effects of voltage drop, coupling terms, and disturbances under nominal parameters. , The disturbance on the dq axis directly reflects parameter mismatch and external disturbances.

[0011] Furthermore, a nonlinear function is designed between the observer estimation error and the perturbation gain, specifically including: Error estimation using a nonlinear function mapping observer With disturbance compensation gain To achieve online adaptive adjustment of the gain, the function is as follows: (11) in, All are adjustable constants, representing the growth rate and saturation characteristics of the equilibrium function.

[0012] Furthermore, a variable-gain Luneburg continuous observer is constructed based on this nonlinear function. The specific process includes: Select the motor dq-axis stator current as the state variable. and the perturbation estimated by the variable gain Romberg continuous observer : The output variable is the actual stator current: The state equations for the motor in the continuous domain are constructed as follows: (12) This includes nominal resistance. Nominal inductance Motor angular velocity System matrix: Control input matrix Output matrix for extracting current components ; The state equations for constructing a variable-gain Romberg continuous observer are as follows: (13) in, The state feedback variable gain matrix of the system, For current tracking gain, This is a nonlinear disturbance compensation gain adjusted based on the observation error e, used to replace the traditional fixed gain. For observation error, The system input state variables are observed by the variable-gain Luneburg continuous observer. This represents the system output variable observed by the variable-gain Romberg continuous observer; For system control variables; The state equations of the variable-gain Romberg continuous observer are converted into discrete form using the forward difference method, and a nonlinear perturbation compensation gain based on the observation error e is introduced. Used to replace traditional fixed gain To balance the system's disturbance rejection and dynamic response speed, the equation for the discrete-domain variable-gain Romberg observer is obtained as follows: (14) coefficient matrix : ; in, , These are the dq-axis stator currents predicted by the variable-gain Luneburg observer in the discrete domain at the next time step. , These represent the dq-axis perturbations predicted by the variable-gain Romberg observer in the discrete domain at the next time step. , Each is the current The dq-axis stator current predicted by a variable-gain Romberg observer in the discrete time domain; , Each is the current dq-axis perturbations predicted by a time-discrete variable-gain Romberg observer; coefficient matrix Includes current tracking gain Nonlinear disturbance compensation gain adjusted based on observation error e and nominal parameter items .

[0013] Furthermore, the stability of the observer in the discrete domain is analyzed using the Julie stability criterion, and the effective range of values ​​for the gain parameter is determined. The specific process includes: Based on the Julie stability criterion in the discrete domain, design the coefficient matrix. The eigenvalues ​​lie within the unit circle to discretize the stability of the system; therefore, the characteristic polynomial must satisfy... After sorting, we get: (15) The characteristic polynomial equation based on the Julie criterion analysis in the discrete domain is used as the discriminant for the stability of the variable-gain Luneburg observer in the discrete domain to determine the current tracking gain. The stable range ensures the convergence of the discrete-domain variable-gain Romberg observer: (16) in, , These are the nominal parameters of the inductor. This represents the inductance mismatch. Sampling time.

[0014] Furthermore, the specific process for generating the optimal reference voltage to compensate for parameter mismatch includes: Using the dq-axis perturbation output of a variable-gain Luneburg observer in the discrete domain, and based on the reference voltage of the hyperlocal model, neglecting the influence of sampling time on small terms, the initial reference voltage form can be obtained: (17) in, The nominal inductance of the motor, , Estimation for a discrete-domain variable-gain Romberg observer The disturbance on the dq axis at time. , For the present Stator current along the dq axis at time t. , This represents the dq-axis stator current at the next moment. The current estimated using the output of a variable-gain Luneburg observer in the discrete domain d-axis perturbation at time t Ignoring the effect of resistance, derive the current... Inductance mismatch at any moment : (18) The inductance mismatch is calculated as follows: (19) The final estimated inductance is the nominal inductance. Subtract the inductor mismatch: (20) Estimating inductance Substitute the nominal inductance into the initial reference voltage form and replace it with the nominal inductance in formula (17). The optimal control voltage for the dq axis to compensate for parameter mismatch is: (twenty one) The input is the dq-axis perturbation estimated by a discrete-domain variable-gain Romberg observer. , The output is the estimated inductance parameters. and the optimal reference voltage along the dq axis at the next moment. , .

[0015] By employing the above technical solution, the present invention provides a current prediction control method based on observer-estimated inductance parameters, which has at least the following beneficial effects: When motor parameters are mismatched, the proposed method has better anti-interference performance. Under inductance mismatch conditions, the proposed method effectively improves the anti-interference capability of the system through inductance estimation. By dynamically adjusting the gain matrix according to the observer error, better compensation for external disturbances is achieved, making the error between the predicted current and the actual current of the dq axis observer significantly smaller than that of the fixed gain method, thus making the current prediction more accurate. This invention addresses the core pain points of Luneburger observer gain adjustment difficulties and strong parameter dependence in current predictive control of permanent magnet synchronous motors, significantly improving system performance through innovative technical solutions. First, a nonlinear function Fun, representing the relationship between observation error and disturbance gain, is designed to replace the traditional fal function, resolving the discrete jitter problem. A variable-gain Luneburger observer is constructed using a hyperlocal model, and the discrete domain stability is ensured by the Julie stability criterion, accelerating observation convergence and enhancing disturbance rejection capability. Second, the inductance parameters are identified online using the observer disturbance quantity, dynamically compensating for parameter mismatches, such as inductance changes caused by temperature rise and magnetic flux coupling, breaking through the dependence of traditional control on precise parameters. Experiments show that with an inductance mismatch of 0.5 or 2.8 times, the proposed method exhibits smaller current tracking error, faster recovery after sudden load increases, and smoother waveforms. This invention is applicable to high-precision drive scenarios such as new energy vehicles and industrial servos, effectively improving the robustness and dynamic response of motor control, and has significant engineering application value. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This invention provides a deadbeat-free current prediction control block diagram based on online adjustment of observer gain according to estimation error; Figure 2 This is a schematic diagram of the curves of the nonlinear function with different parameters and the -fal function proposed in this invention; Figure 3 This is a schematic diagram comparing the phase a current under online gain adjustment and fixed gain when the inductor is at 0.5 times mismatch. Figure 4This is a schematic diagram illustrating the inductance estimation method of the present invention when the inductance is at a mismatch of 0.5 times. Figure 5 This is a schematic diagram illustrating the error between the actual and observed d-axis current values ​​when the inductor is under 0.5 times mismatch, with online adjustment of gain and fixed gain according to the present invention. Figure 6 This is a schematic diagram illustrating the error between the actual and observed q-axis current when the inductor is under 0.5 times mismatch, with online adjustment of gain and fixed gain according to the present invention. Figure 7 This is a schematic diagram comparing the phase a current under online gain adjustment and fixed gain when the inductor is at 2.8 times mismatch. Figure 8 This is a schematic diagram illustrating the inductance estimation of the present invention under an inductance mismatch of 2.8 times; Figure 9 This is a schematic diagram illustrating the error between the actual and observed d-axis current values ​​when the inductor is under 2.8 times mismatch, with online adjustment of gain and fixed gain according to the present invention. Figure 10 This is a schematic diagram illustrating the error between the actual and observed q-axis current when the inductor is under 2.8 times mismatch, with online adjustment of gain and fixed gain according to the present invention. Detailed Implementation

[0017] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. This will allow for a full understanding of how the present application uses technical means to solve technical problems and achieve technical effects, and to facilitate its implementation.

[0018] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0019] Please refer to Figures 1-10 This illustrates a specific implementation of the present embodiment. In the case of motor parameter mismatch, the proposed method has better anti-interference performance. Under the condition of inductance mismatch, the proposed method effectively improves the anti-interference capability of the system through inductance estimation. By dynamically adjusting the gain matrix according to the observer error, better compensation for external disturbances is achieved, making the error between the predicted current and the actual current of the dq axis observer significantly smaller than that of the fixed gain method, thus making the current prediction more accurate.

[0020] Please refer to Figure 1Speed ​​command signal Actual speed signal of permanent magnet synchronous motor (PMSM) Perform deviation calculations to determine the motor's angular velocity. , The number of pole pairs of the motor is given; the resulting speed deviation is input to a PI controller; the PI controller outputs a q-axis current command signal. , and the d-axis current given signal Common input deadbeat current predictive control (DPCC) unit.

[0021] Current coordinate transformation module: Three-phase stator current of the PMSM Input the Clark coordinate transformation cell to convert it into a two-phase stationary state. Current signal in coordinate system and The and The input is the Park coordinate transformation unit, which simultaneously receives the PMSM rotor position signal detected by the encoder. The current signal in a two-phase rotating coordinate system (dq-axis coordinate system) is obtained through Park transformation. and ,and Will With input variable gain Luneburg observer.

[0022] Observation and voltage setting module: The variable gain Luneburger observer outputs the disturbance signal in the dq-axis coordinate system. and Simultaneously, the inductance parameter signal output by the estimated inductor unit, the output perturbation signal of the variable gain Luneburg observer, and the output voltage signal of the DPCC unit are synthesized to obtain the optimal reference voltage setpoint signal in the dq axis coordinate system. and .

[0023] Pulse width modulation and driving module: and The Anti-Park coordinate transformation unit is input, and the Anti-Park unit receives the rotor position signal. Differential signal , converted The voltage signal in the coordinate system is input to the space vector pulse width modulation (SVPWM) unit to generate the inverter switching drive signal.

[0024] Motor and Feedback Module: The switch drive signal is input to the inverter, which converts DC power into three-phase AC power to supply the PMSM; the encoder is mechanically connected to the PMSM rotor to detect the rotor position in real time. .

[0025] This embodiment proposes a current prediction control method based on observer-estimated inductor parameters, which includes the following steps: S1. Construct a discrete current prediction model of the permanent magnet synchronous motor in the dq axis coordinate system, derive the voltage equation for deadbeat current prediction control, and analyze the sensitivity of the motor parameters resistance, inductance and permanent magnet flux mismatch to current tracking performance, and obtain the control voltage for parameter mismatch. As a preferred embodiment of step S1, the specific process includes the following steps: S11. Establish the dynamic equations for the stator current of the permanent magnet synchronous motor: (1) Matrix definition: , , State variables: For the dq-axis stator current, the input variable is: This refers to the dq-axis stator voltage. For the d-axis stator inductance, For the q-axis stator inductance, Let be the angular velocity of the motor. R is the actual magnetic flux linkage of the rotor permanent magnet during operation, and R is the actual resistance of the motor during operation. For general label-type motors, through The simplified matrix is ​​obtained by simplifying the surface-mount motor, where L is the inductance parameter of the motor during actual operation.

[0026] S12. Based on formula (1) and using Euler discretization, when the sampling time... When it is long enough, according to the current Predicting stator current information at the next time step by using stator current at a given time step transforms the continuous model into a discrete current prediction model: (2) The simplified matrix after discretization is: , , in, The dq-axis stator current at the next moment. For the present The dq-axis stator current at time t. Sampling time.

[0027] S13. By selecting the inverter output voltage, the dq-axis stator current at the next moment is made... Accurately track the reference current in the next sampling period ,Right now: (3) current Stator voltage along the dq axis at time 1 The reference current of the dq axis at the next moment. , Substituting the next moment's dq-axis stator current into formula (2), the optimal control voltage equation in the deadbeat current predictive control (DPCC) of the permanent magnet synchronous motor (PMSM) under parameter mismatch is calculated as follows: (4) in, , , The corresponding discrete matrix; through the current The optimal control voltage required to predict the stator current at the next moment is determined by the stator current at the current moment, so as to achieve the control objective of making the stator current at the next moment accurately track the reference current. Based on the surface-mount motor unfolded into the specific form of the dq axis, we obtain , As a prediction of the present Optimal control voltage is maintained at all times to control the inverter's output power and drive the motor to track the reference current. (5) in, , This is the resistance voltage drop term. , for The rate of change of current in the inductor required to track the reference current during the sampling time. , For dq axis cross-coupling terms, This is the back electromotive force term of the permanent magnet.

[0028] S14. During motor operation, phenomena such as motor temperature rise and magnetic flux coupling may occur, causing changes in rated parameters such as inductance. This leads to a mismatch between the parameters of the control inverter and the actual parameters during motor operation. As can be seen from equation (1), the dynamic equation of the stator current of a permanent magnet synchronous motor includes three important parameters: resistance, inductance, and magnetic flux. When the motor parameters are mismatched, there will be an error between the actual stator current and the reference current, meaning that the dq-axis stator current cannot be adjusted at the next moment. Follow the reference current This could also lead to system instability. During stable motor operation, parameter mismatch can cause the voltage to deviate from its optimal value, affecting current tracking accuracy. , , For formula (1) and The control voltage equation for the parameter mismatch is calculated as follows: (6) in, , , These are the nominal parameters of the motor's resistance, inductance, and flux linkage. , , This refers to the mismatch in resistance, inductance, and flux linkage during motor operation.

[0029] S2. Define the parameter mismatch and external disturbances during motor operation as a unified disturbance quantity. Based on the hyperlocal model, the control voltage equation for parameter mismatch is transformed into a current dynamic equation; As a preferred embodiment of step S2, the specific process includes the following steps: S21. The parameter mismatches of resistance, inductance, and flux linkage during motor operation, along with external interference, are aggregated into a unified disturbance. To simplify the observer design, the perturbation set is defined as follows: (7) in, This represents the disturbance along the dq axis. This represents the perturbation term due to cross-coupling and flux variation; , , This refers to the mismatch in resistance, inductance, and flux linkage during motor operation. For the dq axis stator current, Let be the derivative of the dq-axis stator current. This represents the angular velocity of the motor. The control voltage equation after parameter mismatch is transformed into: (8) in, , , For the corresponding discrete matrix, the perturbation set definition transforms multi-source disturbances into observable vectors. This provides a unified objective for observers to estimate disturbances; S22, Based on Hyperlocal Model ,in The gain of the dq-axis control voltage. Let dq-axis disturbance be represented. A direct relationship is established between the rate of change of current and the control voltage and disturbance. The rate of change of current is a linear combination of the disturbance and the control voltage. The control voltage equation (8) of the transformed parameter mismatch is converted into the current dynamic equation, specifically in the form of the current derivative: (9) in, , To control the contribution of voltage to changes in nominal inductance parameters, The combined effects of voltage drop, coupling terms, and disturbances under nominal parameters. , The disturbance on the dq axis directly reflects parameter mismatch and external disturbances.

[0030] S3. Design a nonlinear function between the observer estimation error and the perturbation gain. Based on this nonlinear function, construct a variable-gain Romberg continuous observer. Analyze the stability of the observer in the discrete domain using the Julie stability criterion and determine the effective range of the gain parameter. As a preferred embodiment of step S3, the specific process includes the following steps: To estimate the disturbance using a variable-gain Luneburg continuous observer, the gain term of the observer needs to be designed. Gain term It usually consists of two parts: and .in, The weight corresponding to the current prediction error is mainly used to enhance current tracking accuracy and reduce the deviation between the actual current and the predicted current; its value is relatively easy to determine. This corresponds to the disturbance compensation gain, primarily used to enhance the system's disturbance rejection capability and improve robustness. During motor operation, factors such as model parameter disturbances, external load changes, and speed switching can all cause interference to the system, thus requiring the observer to have strong disturbance rejection capabilities. Typically, The larger the value, the higher the system's tolerance to parameter changes and external disturbances; however, if... If the current is too large, it will weaken the system's sensitivity to tracking current commands, resulting in a slower dynamic response, or even sacrificing control accuracy due to excessive "suppression". Therefore, a nonlinear function is used to map the nonlinear relationship between the observer estimation error and the disturbance gain.

[0031] In addressing the problem of nonlinear processing of error signals, researchers proposed the fal function, as shown in the following equation: (10) When adjusting the observer gain using the fal function, the eigenvalues ​​of the observer system equations are difficult to obtain analytically. Furthermore, the fal function is not differentiable at piecewise points, which can easily cause numerical jitter during discrete implementation. When the observation error is large, the fal function exhibits power-law growth, potentially leading to overcompensation or even system oscillation. Therefore, this invention proposes a nonlinear gain function suitable for discrete systems.

[0032] S31. This invention uses a nonlinear function mapping observer to estimate the error. With disturbance compensation gain To achieve online adaptive adjustment of the gain, the function is as follows: (11) in, All are adjustable constants, representing the growth rate and saturation characteristics of the equilibrium function.

[0033] It should be noted that the fitting curves of the Fun function and the fal function are as follows: Figure 2 As shown: the horizontal axis represents the observer estimation error e, and the vertical axis represents the disturbance compensation gain, containing multiple sets of different parameters. nonlinear functions The curve, and the -fal function curve, The function curve has a high degree of fit with the -fal function curve, which verifies that the proposed nonlinear function can effectively replace the fal function, solve the jitter defect in its discrete implementation, and provide a reliable functional form for the dynamic gain adjustment of the observer.

[0034] Since discrete systems struggle to handle complex nonlinear functions, the derivative of the aforementioned function is approximated to obtain... The approximate function is directly used as the perturbation compensation gain in the gain term K of the variable gain Romberg continuous observer. In the form of, that is, using to replace This function has the characteristics of large gain (strong disturbance rejection) when the error is large and small gain (fast response) when the error is small. It is also continuously differentiable and jitter-free, making it suitable for discrete observer implementation. The approximate nonlinear function solves the defects of the fal function and provides a practical gain adjustment method for the subsequent construction of variable gain Romberg continuous observer. In summary, by using disturbance concentration, model transformation, gain design, and function approximation, multi-source disturbances are transformed into observable disturbances. An adaptive nonlinear gain adjustment method is designed, and the core theoretical framework of the variable-gain Luneburg continuous observer is constructed, which is the foundation for achieving high-precision current prediction and parameter identification.

[0035] S32. Select the state variable as the motor dq-axis stator current. and the perturbation estimated by the variable gain Romberg continuous observer : The output variable is the actual stator current: The state equations for the motor in the continuous domain are constructed as follows: (12) This includes nominal resistance. Nominal inductance Motor angular velocity System matrix: Control input matrix Output matrix for extracting current components ; The state equations for constructing a variable-gain Romberg continuous observer are as follows: (13) in, The state feedback variable gain matrix of the system, For current tracking gain, This is a nonlinear disturbance compensation gain adjusted based on the observation error e, used to replace the traditional fixed gain. , For observation error, The system input state variables are observed by the variable-gain Luneburg continuous observer. This represents the system output variable observed by the variable-gain Romberg continuous observer; These are the system control variables; by constructing an observation model parallel to the original system, the estimated state is corrected using output error feedback, thereby achieving real-time estimation of the internal current and disturbance states of the system. S33. Since the actual control system is a discrete sampling system, the forward difference method is needed to convert the state equation of the variable gain Romberg continuous observer into a discrete form (using the forward difference method, due to the sampling time...). (Sufficiently small); introduce a nonlinear perturbation gain adjusted based on the observation error e. Used to replace traditional fixed gain Balancing the system's disturbance rejection and dynamic response speed, the equation for the discrete-domain variable-gain Romberg observer is obtained as follows: (14) coefficient matrix : ; in, , These are the dq-axis stator currents predicted by the variable-gain Luneburg observer in the discrete domain at the next time step. , These represent the dq-axis perturbations predicted by the variable-gain Romberg observer in the discrete domain at the next time step. , Each is the current The dq-axis stator current predicted by a variable-gain Romberg observer in the discrete time domain; , Each is the current dq-axis perturbation predicted by a variable-gain Romberg observer in the discrete time domain.

[0036] coefficient matrix Includes current tracking gain Nonlinear disturbance compensation gain adjusted based on observation error e and nominal parameter items , is the core parameter of the discrete-domain variable-gain Luneburg observer, directly affecting the observer's performance and stability: matrix D is the core part of the discrete-domain variable-gain Luneburg observer in Equation 14, directly determining the observer's state transition dynamics. The eigenvalues ​​of the matrix must satisfy the stability condition (Julie criterion) to ensure the convergence of the observer; therefore, their calculation is a key step in the implementation of Equation 14.

[0037] S34, Coefficient Matrix The eigenvalues ​​determine the rate of decay of the estimation error. To stabilize the variable-gain Romberg observer in the discrete domain, the coefficient matrix in equation (14) The eigenvalues ​​must all be negative. Based on the Julie stability criterion for the discrete domain, the coefficient matrix is ​​designed. The eigenvalues ​​lie within the unit circle to discretize the stability of the system; therefore, the characteristic polynomial must satisfy... After sorting, we get: (15) The characteristic polynomial equation based on the Julie criterion analysis in the discrete domain is used as the discriminant for the stability of the variable-gain Luneburg observer in the discrete domain to determine the current tracking gain. The stable range ensures the convergence of the discrete-domain variable-gain Romberg observer: (16) in, The nonlinear perturbation gain is adjusted based on the observation error e. , These are the nominal parameters of the inductor. This represents the inductance mismatch. Sampling time; select within this range This ensures stable operation of the discrete-domain variable-gain Luneburg observer, accurately estimates disturbances, and compensates for the voltage vector of optimal control.

[0038] S4. The disturbance output of the discrete domain variable gain Luneburg observer is used to identify the motor inductance parameters in real time. The identified estimated inductance parameters are substituted into the initial reference voltage calculation formula to generate the optimal reference voltage to compensate for parameter mismatch, thereby realizing high-precision current prediction control of permanent magnet synchronous motor. As a preferred embodiment of step S4, the specific process includes the following steps: S41, due to sampling time Small enough that its effect on the nominal parameter items is negligible. and motor angular velocity The influence of the variable gain Luneburger observer in the discrete domain is utilized to measure the dq-axis perturbation of the observer output. Based on the reference voltage of the hyperlocal model, the sampling time is ignored. The effect of the minor terms can be used to obtain the form of the initial reference voltage: (17) in, The nominal inductance of the motor, , Estimation for a discrete-domain variable-gain Romberg observer The disturbance on the dq axis at time. , For the present Stator current along the dq axis at time t. , This represents the dq-axis stator current at the next moment. Parameter stability analysis shows that the inductance parameter affects the range of the gain term of the variable-gain Luneburg observer in the discrete domain, and also affects the system stability. Therefore, this application proposes to implement dynamic perturbation gain based on the observation error of the variable-gain Luneburg observer in the discrete domain, and then update the inductance parameter through the perturbation. The control architecture diagram is shown below. Figure 1 As shown.

[0039] S42. Estimating the current output using a variable-gain Roenberger observer in the discrete domain. d-axis perturbation at time t Ignoring the effect of resistance, derive the current... Inductance mismatch at any moment : (18) The inductance mismatch is calculated as follows: (19) The final estimated inductance is the nominal inductance. Subtract the inductor mismatch: (20) S43, Estimate Inductance Substitute the nominal inductance into the initial reference voltage form and replace it with the nominal inductance in formula (17). The optimal control voltage for the dq axis to compensate for parameter mismatch is: (twenty one) The input is the dq-axis perturbation estimated by a discrete-domain variable-gain Romberg observer. , The output is the estimated inductance parameters. and the optimal reference voltage along the dq axis at the next moment. , .

[0040] In this embodiment, the design of the variable-gain Luenberger observer is fundamental. Stability analysis ensures the reliable operation of the discrete-domain variable-gain Luenberger observer, providing accurate disturbance quantities for subsequent inductor identification. A stable discrete-domain variable-gain Luenberger observer outputs reliable disturbance quantities, which is a prerequisite for inductor identification. The estimated inductance and optimal reference voltage are fed back to the current prediction control loop to compensate for parameter mismatch, forming a closed-loop optimization of observation, stabilization, and compensation. Specifically, step S43, generating the optimal dq-axis control voltage to compensate for parameter mismatch, relies on the estimated disturbance quantity output by the discrete-domain variable-gain Luenberger observer in step S33 to ensure the accuracy of inductor identification, and the construction of a stable discrete-domain variable-gain Luenberger observer in step S34 to ensure the estimated disturbance quantity is without deviation. Simultaneously, the stator current data from step S1 is required. , In step S43, the optimal dq-axis control voltage generated by the compensation parameter mismatch is directly fed back to the current prediction control loop, which solves the problem of decreased control accuracy caused by motor parameter mismatch (such as inductance change) and improves system robustness; the estimated inductance parameters can also be used to update the observer model to form closed-loop optimization.

[0041] In summary, step S43 serves as a bridge between the discrete-domain variable-gain Luenberger observer and actual control. By dynamically estimating the inductor parameters and adjusting the optimal reference voltage, parameter mismatch is effectively compensated, achieving high-precision current predictive control. Inductor identification solves the parameter mismatch problem, and the generation of the optimal reference voltage directly improves current tracking accuracy and system robustness.

[0042] The following are specific embodiments of the present invention: The motor parameters of this invention are shown in Table 1. The motor starts with a load of 0.3 N·m and a speed of 900 rpm, and the load is suddenly increased to 0.8 N·m at 0.5 s, as shown in the table below:

[0043] Depend on Figures 3 to 10 As can be seen from the comparison, Figure 3This diagram illustrates the comparison of phase a current under online gain adjustment and fixed gain when the inductor is under 0.5 times mismatch. The horizontal axis represents time (s), and the vertical axis represents the phase a current value. (A) The current waveform of phase a under the online gain adjustment strategy is more stable than that under the fixed gain strategy. It recovers faster and has a smaller tracking error after a sudden load increase (from 0.3 N·m to 0.8 N·m in 0.5 s). The online gain adjustment strategy can better maintain current stability and improve system immunity when there is a small inductor mismatch. Figure 4 The diagram shows the inductance estimation under a 0.5 mismatch condition. The horizontal axis represents time (s), and the vertical axis represents the estimated inductance value. The estimated inductance converges quickly to the actual inductance value with a small steady-state error, which verifies the effectiveness of the inductance identification algorithm. It can track parameter changes in real time and provide an accurate basis for parameter compensation. Figure 5 This diagram illustrates the comparison of d-axis current observation errors under 0.5x inductor mismatch conditions. The horizontal axis represents time (s), and the vertical axis represents the error between the actual and observed d-axis current values. (A) The error amplitude of online gain adjustment is smaller, the convergence is faster, and the steady-state error is almost zero. Online gain adjustment significantly improves the accuracy of d-axis current observation and reduces the impact of parameter mismatch on observation. Figure 6 This diagram illustrates the comparison of q-axis current observation errors under an inductance mismatch of 0.5. The horizontal axis represents time (s), and the vertical axis represents the error between the actual and observed q-axis current values. (A) Online gain adjustment results in smaller errors and smaller fluctuations after sudden load application. Online gain adjustment also improves the accuracy of q-axis current observation and ensures the accuracy of torque control. Figure 7 This diagram illustrates the comparison between the phase a current under online gain adjustment and fixed gain with an inductance mismatch of 2.8 times. The horizontal axis represents time (s), and the vertical axis represents the phase a current value. (A), same Figure 3 However, the inductor mismatch is larger (2.8 times the nominal value), and the current remains stable when the gain is adjusted online. It recovers quickly after a sudden load is applied. The online gain adjustment strategy can still maintain current stability under large inductor mismatch, which is extremely robust. Figure 8 The diagram shows the inductance estimation under a 2.8 times inductance mismatch. The horizontal axis represents time (s) and the vertical axis represents the estimated inductance value. The estimated inductance converges to the actual inductance value. The convergence speed is fast and the steady-state error is small. The inductance identification algorithm is still effective under large parameter mismatch, which verifies the adaptability of the algorithm. Figure 9 The comparison of d-axis current observation errors under an inductance mismatch of 2.8 times is shown, with the horizontal axis representing time (s) and the vertical axis representing the d-phase current error. (A), same Figure 5However, the inductor mismatch is larger, and the error amplitude under the fixed gain strategy is significantly larger than that under the online gain adjustment strategy. The online gain adjustment converges faster, and the online gain adjustment can still ensure the accuracy of d-axis current observation under large mismatch, thus avoiding system instability. Figure 10 The comparison shows the q-axis current observation error under an inductance mismatch of 2.8 times. The horizontal axis represents time (s), and the vertical axis represents the q-phase current error. (A) Under the fixed gain strategy, the error fluctuation is severe. The error of online gain adjustment is small and the steady-state error is small. Online gain adjustment can still maintain the accuracy of q-axis current observation under extreme parameter mismatch, ensuring stable torque output.

[0044] In summary, the proposed method exhibits better anti-interference performance when motor parameters are mismatched. Under inductance mismatch conditions, the proposed method effectively improves the system's anti-interference capability by estimating the inductance. By dynamically adjusting the gain matrix based on the error of the variable-gain Luenberger observer in the discrete domain, it achieves better compensation for external disturbances. This makes the error between the current predicted by the variable-gain Luenberger observer in the dq-axis discrete domain and the actual current significantly smaller than that of the fixed-gain method, resulting in more accurate current prediction.

[0045] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus or device (such as a computer-based system, a processor-included system or other system that can fetch and execute instructions from, an instruction execution system, apparatus or device).

[0046] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A current prediction control method based on observer-estimated inductance parameters, characterized in that, Includes the following steps: A discrete current prediction model of a permanent magnet synchronous motor in the dq-axis coordinate system is constructed, the voltage equation for deadbeat current prediction control is derived, and the sensitivity of the motor parameters resistance, inductance and permanent magnet flux mismatch to current tracking performance is analyzed to obtain the control voltage for parameter mismatch. The parameter mismatch and external disturbances during motor operation are collectively defined as a unified disturbance. Based on the hyperlocal model, the control voltage equation for parameter mismatch is transformed into a current dynamic equation; Design a nonlinear function relating the observer estimation error to the perturbation gain. Construct a variable-gain Romberg continuous observer based on this nonlinear function. Analyze the stability of the observer in the discrete domain using the Julie stability criterion and determine the effective range of values ​​for the gain parameter. The inductance parameters of the motor are identified in real time by using the disturbance output of the discrete domain variable gain Luneburg observer. The estimated inductance parameters obtained by identification are substituted into the initial reference voltage calculation formula to generate the optimal reference voltage to compensate for parameter mismatch, thereby realizing high-precision current predictive control of permanent magnet synchronous motor.

2. The current prediction control method based on observer-estimated inductance parameters according to claim 1, characterized in that: The process of constructing a discrete current prediction model for a permanent magnet synchronous motor in the dq-axis coordinate system includes: Establish the dynamic equation for the stator current of the permanent magnet synchronous motor: (1) Matrix definition: , , State variables: For the dq-axis stator current, the input variable is: This refers to the dq-axis stator voltage. For the d-axis stator inductance, For the q-axis stator inductance, Let be the angular velocity of the motor. R is the actual magnetic flux linkage of the rotor permanent magnet during operation, and R is the actual resistance of the motor during operation. pass A simplified matrix is ​​obtained by simplifying the surface-mount motor, where L represents the inductance parameter of the motor during actual operation, and Euler discretization is used, based on the current... Predicting stator current information at the next time step by using stator current at a given time step transforms the continuous model into a discrete current prediction model: (2) The simplified matrix after discretization is: , , in, The dq-axis stator current at the next moment. For the present The dq-axis stator current at time t.

3. The current prediction control method based on observer-estimated inductance parameters according to claim 2, characterized in that: The specific calculation process for obtaining the control voltage for parameter mismatch includes: By selecting the inverter output voltage, the dq-axis stator current at the next moment is... Accurately track the reference current in the next sampling period ,Right now: (3) current Stator voltage along the dq axis at time 1 The reference current of the dq axis at the next moment. , Substituting the next moment's dq-axis stator current into formula (2), the optimal control voltage equation in the deadbeat current predictive control (DPCC) of the permanent magnet synchronous motor (PMSM) under parameter mismatch is calculated as follows: (4) in, , , The corresponding discrete matrix; Through the current The optimal control voltage required to predict the stator current at the next moment is obtained based on the surface-mounted motor unfolded into a dq-axis configuration. , As a prediction of the present Optimal control voltage is maintained at all times to control the inverter's output power and drive the motor to track the reference current. （5） in, , This is the resistance voltage drop term. , for The rate of change of current in the inductor required to track the reference current during the sampling time. , For dq axis cross-coupling terms, This is the back electromotive force term of the permanent magnet; During stable motor operation, when parameter mismatch occurs: , , The control voltage equation for the parameter mismatch is calculated as follows: （6） in, , , These are the nominal parameters of the motor's resistance, inductance, and flux linkage. , , This refers to the mismatch in resistance, inductance, and flux linkage during motor operation.

4. The current prediction control method based on observer-estimated inductance parameters according to claim 1, characterized in that: The parameter mismatch and external disturbances during motor operation are collectively defined as a unified disturbance. Specifically, it includes: The parameter mismatches of resistance, inductance, and flux linkage during motor operation, along with external disturbances, are aggregated into a unified disturbance. To simplify the observer design, the perturbation set is defined as follows: (7) in, This represents the disturbance along the dq axis. This represents the perturbation term due to cross-coupling and flux variation; , , This refers to the mismatch in resistance, inductance, and flux linkage during motor operation. For the dq axis stator current, Let be the derivative of the dq-axis stator current. This represents the angular velocity of the motor. The control voltage equation after parameter mismatch is transformed into: (8) in, , , is the corresponding discrete matrix.

5. The current prediction control method based on observer-estimated inductance parameters according to claim 4, characterized in that: Based on the hyperlocal model, the control voltage equation for parameter mismatch is transformed into a current dynamic equation, specifically including: Based on hyperlocal model ,in The gain of the dq-axis control voltage. Let dq-axis disturbance be represented. A direct relationship is established between the rate of change of current and the control voltage and disturbance. The rate of change of current is a linear combination of the disturbance and the control voltage. The control voltage equation (8) of the transformed parameter mismatch is converted into the current dynamic equation, specifically in the form of the current derivative: （9） in, , To control the contribution of voltage to changes in nominal inductance parameters, The combined effects of voltage drop, coupling terms, and disturbances under nominal parameters. , The disturbance on the dq axis directly reflects parameter mismatch and external disturbances.

6. The current prediction control method based on observer-estimated inductance parameters according to claim 3, characterized in that: Design the nonlinear function between the observer estimation error and the perturbation gain, specifically including: Error estimation using a nonlinear function mapping observer With disturbance compensation gain To achieve online adaptive adjustment of the gain, the function is as follows: （11） in, All are adjustable constants, representing the growth rate and saturation characteristics of the equilibrium function.

7. The current prediction control method based on observer-estimated inductance parameters according to claim 6, characterized in that: The variable-gain Romberg continuous observer is constructed based on this nonlinear function. The specific process includes: Select the motor dq-axis stator current as the state variable. and the perturbation estimated by the variable gain Romberg continuous observer : The output variable is the actual stator current: The state equations for the motor in the continuous domain are constructed as follows: （12） This includes nominal resistance. Nominal inductance Motor angular velocity System matrix: Control input matrix Output matrix for extracting current components ; The state equations for constructing a variable-gain Romberg continuous observer are as follows: (13) in, The state feedback variable gain matrix of the system, For current tracking gain, This is a nonlinear disturbance compensation gain adjusted based on the observation error e, used to replace the traditional fixed gain. For observation error, The system input state variables are observed by the variable-gain Luneburg continuous observer. This represents the system output variable observed by the variable-gain Romberg continuous observer; For system control variables; The state equations of the variable-gain Romberg continuous observer are converted into discrete form using the forward difference method, and a nonlinear perturbation compensation gain based on the observation error e is introduced. Used to replace traditional fixed gain To balance the system's disturbance rejection and dynamic response speed, the equation for the discrete-domain variable-gain Romberg observer is obtained as follows: （14） coefficient matrix : ； in, , These are the dq-axis stator currents predicted by the variable-gain Luneburg observer in the discrete domain at the next time step. , These represent the dq-axis perturbations predicted by the variable-gain Romberg observer in the discrete domain at the next time step. , Each is the current The dq-axis stator current predicted by a variable-gain Romberg observer in the discrete time domain; , Each is the current dq-axis perturbations predicted by a time-discrete variable-gain Romberg observer; coefficient matrix Includes current tracking gain Nonlinear disturbance compensation gain adjusted based on observation error e and nominal parameter items .

8. The current prediction control method based on observer-estimated inductance parameters according to claim 7, characterized in that: The stability of the observer in the discrete domain is analyzed using the Julie stability criterion, and the effective range of values ​​for the gain parameter is determined. The specific process includes: Based on the Julie stability criterion in the discrete domain, design the coefficient matrix. The eigenvalues ​​lie within the unit circle to discretize the stability of the system; therefore, the characteristic polynomial must satisfy... After sorting, we get: （15） The characteristic polynomial equation based on the Julie criterion analysis in the discrete domain is used as the discriminant for the stability of the variable-gain Luneburg observer in the discrete domain to determine the current tracking gain. The stable range ensures the convergence of the discrete-domain variable-gain Romberg observer: （16） in, , These are the nominal parameters of the inductor. This represents the inductance mismatch. Sampling time.

9. The current prediction control method based on observer-estimated inductance parameters according to claim 8, characterized in that: The specific process for generating the optimal reference voltage to compensate for parameter mismatch includes: Using the dq-axis perturbation output of a variable-gain Luneburg observer in the discrete domain, and based on the reference voltage of the hyperlocal model, neglecting the influence of sampling time on small terms, the initial reference voltage form can be obtained: （17） in, The nominal inductance of the motor, , Estimation for a discrete-domain variable-gain Romberg observer The disturbance on the dq axis at time. , For the present Stator current along the dq axis at time t. , This represents the dq-axis stator current at the next moment. The current estimated using the output of a variable-gain Luneburg observer in the discrete domain d-axis perturbation at time t Ignoring the effect of resistance, derive the current... Inductance mismatch at any moment : （18） The inductance mismatch is calculated as follows: （19） The final estimated inductance is the nominal inductance. Subtract the inductor mismatch: （20） Estimating inductance Substitute the nominal inductance into the initial reference voltage form and replace it with the nominal inductance in formula (17). The optimal control voltage for the dq axis to compensate for parameter mismatch is: （21） The input is the dq-axis perturbation estimated by a discrete-domain variable-gain Romberg observer. , The output is the estimated inductance parameters. and the optimal reference voltage along the dq axis at the next moment. , .

Images