Permanent magnet synchronous motor identification device and method based on source reconstruction and instruction presetting

Abstract

The application discloses a kind of permanent magnet synchronous motor identification device and method of source reconstruction and instruction preset coordination, suitable for the vector control working condition of d-axis given current is zero.The device includes source reconstruction unit, instruction preset unit, flux linkage domain MRAS identification unit and gain scheduling unit.Source reconstruction unit carries out closed-loop calibration to encoder original electric angle, and output calibration electric angular velocity and internal electric angle;Instruction preset unit generates q-axis feedforward current according to speed instruction, and is superimposed with speed loop PI output to form q-axis given current;Flux linkage domain MRAS identification unit constructs reference model and adjustable model, and gain scheduling unit updates inductance reciprocal and flux linkage inductance ratio according to the flux linkage error between reference flux linkage and adjustable flux linkage, calculates equivalent inductance and equivalent permanent magnet flux linkage, and updates current loop PI parameter.The application can reduce angular velocity noise and dynamic current tracking error influence, improve identification stability and control performance.

Description

Technical Field

[0001] This invention belongs to the field of motor control technology and relates to an online identification device and method for permanent magnet synchronous motor parameters. In particular, it relates to a device and method for identifying equivalent inductance and equivalent permanent magnet flux linkage parameters online by improving the quality of the MRAS error signal in the flux linkage domain through source reconstruction and command preset coordination under vector control conditions where the d-axis given current is zero. Background Technology

[0002] Permanent magnet synchronous motors (PMSMs) have advantages such as high power density, high efficiency, and fast response, and are widely used in electric vehicles, industrial servo systems, and robotics. However, during operation, the internal parameters of PMSMs change. For example, increased temperature can lead to a decrease in the magnetic flux linkage of the permanent magnets, increased current can cause magnetic circuit saturation, and this can result in a decrease in the equivalent inductance.

[0003] In vector control systems, the parameter design of the current loop controller depends on the accurate values ​​of inductance and flux linkage. When the parameters used by the controller do not match the actual parameters of the motor, problems such as decreased current tracking performance, increased torque ripple, and reduced efficiency will occur. Therefore, real-time identification of these two key parameters, inductance and flux linkage, during motor operation is crucial for ensuring control performance.

[0004] Model Reference Adaptive Systems (MRAS) are widely used for online identification of motor parameters due to their clear structure and well-defined physical meaning. The basic idea is to establish a reference model independent of the parameters to be identified, and then establish an adjustable model that includes the parameters. The differences in their outputs are compared, and the parameters in the adjustable model are continuously corrected using an adaptive law based on the error signal, until the outputs of the two models tend to be consistent. At this point, the parameters in the adjustable model represent the true values ​​of the actual parameters.

[0005] However, traditional MRAS faces three main problems in practical applications:

[0006] First, inconsistencies between encoder quantization noise and angle measurement introduce errors. The angle signal output by the encoder contains quantization steps and glitches, and the speed signal obtained after differentiation has severe high-frequency noise. These noisy signals are injected into the error signal construction process of MRAS through coordinate transformation, resulting in obvious transient spikes and steady-state oscillations in the parameter identification results.

[0007] Second, during the rapid acceleration and deceleration of the motor, the current command output by the speed loop controller exhibits dynamic lag, resulting in a significant tracking error between the actual current and the commanded current. This dynamic current error contaminates the flux linkage calculation of the reference model, causing the error signal of MRAS to contain large-energy disturbances during transient processes, severely affecting the convergence speed and stability of parameter identification.

[0008] Third, while existing technologies include methods such as speed filtering, phase-locked loop calibration, torque feedforward compensation, and MRAS parameter identification, these methods typically operate independently. Speed ​​filtering is mainly used for speed feedback smoothing, torque feedforward is mainly used to improve speed tracking, and MRAS is mainly used for parameter estimation. Existing solutions rarely address the formation mechanism of MRAS error signals by unifying the speed signal source purification and current command source pre-compensation at the front end of the flux linkage domain MRAS.

[0009] More specifically, encoder noise enters the reference flux calculation through coordinate transformation and the rotating back EMF term, while dynamic current tracking error enters the comparison stage between the adjustable flux and the reference flux through voltage commands and current feedback. These two factors create cross-interference at the MRAS error signal construction point. Simply filtering the encoder signal is insufficient to suppress the dynamic current error during rapid acceleration and deceleration; adding only torque feedforward is insufficient to eliminate the impact of angular velocity estimation noise on the reference model.

[0010] Therefore, how to simultaneously improve the quality of the MRAS error signal in the flux linkage domain from both the signal source and the command source, and improve the stability, convergence speed, and controller parameter matching of equivalent inductance and equivalent permanent magnet flux linkage identification while ensuring engineering feasibility, is a technical problem that needs to be solved. In the existing technology, even if angle filtering or phase-locked loop calibration is used alone, it is mainly used to reduce the influence of encoder quantization noise and angular velocity differential glitch on coordinate transformation and the back EMF term of the reference model rotation, but it is still difficult to eliminate the dynamic current error caused by current command lag during rapid acceleration and deceleration; while using q-axis feedforward compensation alone can provide acceleration torque in advance and reduce current tracking lag, it cannot suppress high-frequency noise in the encoder's original angle and its differential velocity. Therefore, if the two are used as independent modules, their improvement targets and action links are separated, making it difficult to simultaneously reduce angular velocity noise and dynamic current error from the source of MRAS error signal formation. This invention uses a signal source reconstruction unit ( Output The calibration of the electric angular velocity and the q-axis given current output by the command preset unit are jointly introduced into the magnetic flux domain MRAS reference model and error signal construction process, so that the angular velocity noise and dynamic current error are synchronously weakened before entering the parameter adaptive law, thus forming a synergistic suppression effect at the source of MRAS error signal, rather than a simple parallel process of angle filtering, q-axis feedforward compensation and MRAS identification. Summary of the Invention

[0011] The technical problem solved by this invention is to provide a permanent magnet synchronous motor identification device and method that combines source reconstruction and command preset. The device performs closed-loop calibration of the encoder signal through the source reconstruction mechanism and performs advance compensation of dynamic current through the command preset mechanism. The two mechanisms work together to improve the quality of MRAS error signal from the signal source and command source, respectively, thereby improving the stability and convergence performance of parameter identification.

[0012] This invention is achieved through the following technical solution:

[0013] A permanent magnet synchronous motor identification device with coordinated signal source reconstruction and command preset includes a signal source reconstruction unit, a command preset unit, a flux linkage domain MRAS identification unit, and a gain scheduling unit. The input terminal of the signal source reconstruction unit is connected to an encoder to receive the raw electrical angle output by the encoder. The output terminal of the signal source reconstruction unit is connected to the flux linkage domain MRAS identification unit to output... Calibrate the electric angular velocity; the The calibrated electric angular velocity is converted by the pole pair conversion module. After calibrating the mechanical angular velocity, the input command preset unit is used. The input of the command preset unit is also connected to a speed command source, and the output of the command preset unit is connected to a flux linkage domain MRAS identification unit. This unit provides d-axis and q-axis setpoint currents to the current loop PI controller within the flux linkage domain MRAS identification unit. The q-axis setpoint current is formed by superimposing the q-axis feedforward current and the speed loop PI output. The d-axis setpoint current is set to zero under vector control conditions where the d-axis setpoint current is zero. The flux linkage domain MRAS identification unit includes a current loop PI controller, a reference model, and an adjustable model. The current loop PI controller outputs d-axis and q-axis voltage commands based on the d-axis setpoint current, the q-axis setpoint current, and the d- and q-axis current feedback values. The reference model outputs d-axis and q-axis voltage commands based on... The reference flux linkage is reconstructed using calibrated electric angular velocity, d-axis voltage command, q-axis voltage command, and d- and q-axis current feedback values. The adjustable model constructs d- and q-axis adjustable flux linkages based on the inverse inductance estimate, flux linkage-to-inductance ratio estimate, and d- and q-axis current feedback values. The output of the flux linkage domain MRAS identification unit is connected to the gain scheduling unit, providing the reference flux linkage and adjustable flux linkage. The gain scheduling unit updates the inverse inductance estimate and flux linkage-to-inductance ratio estimate based on the flux linkage error between the reference and adjustable flux linkages, and updates the parameters of the current loop PI controller within the flux linkage domain MRAS identification unit based on the updated inductance and flux linkage-to-inductance ratio estimates.

[0014] The source reconstruction unit is used to perform closed-loop calibration of the raw electrical angle output by the encoder. Calibrate electric angular velocity and Internal electrical angle, and according to The electric angular velocity was obtained by calibration. The system calibrates the mechanical angular velocity to suppress measurement noise interference in subsequent identification processes from the signal source side; a command preset unit is used to pre-generate the q-axis feedforward current based on the speed command and mechanical motion equations, and superimposes the q-axis feedforward current with the speed loop PI output to form the q-axis setpoint current. Simultaneously, it provides the d-axis setpoint current under vector control conditions where the d-axis setpoint current is zero, reducing current tracking error during dynamic processes from the current command source side; a flux linkage domain MRAS identification unit is used to... A reference model is constructed using calibrated electric angular velocity, d-axis voltage command, q-axis voltage command, and d- and q-axis current feedback values ​​to obtain a reference flux linkage. An adjustable model is then constructed based on the inverse inductance estimate, flux linkage-to-inductance ratio estimate, and d- and q-axis current feedback values ​​to obtain an adjustable flux linkage. A gain scheduling unit, connected to the flux linkage domain MRAS identification unit, generates an instantaneous drive signal based on the flux linkage error between the reference and adjustable flux linkages. This instantaneous drive signal is dynamically shaped via a linear forward channel to obtain a shaped drive signal. The inverse inductance estimate and flux linkage-to-inductance ratio estimate are updated using a PI adaptive law. Finally, the parameters of the current loop PI controller within the flux linkage domain MRAS identification unit are updated based on these updated inductance and flux linkage-to-inductance ratio estimates.

[0015] The aforementioned source reconstruction unit and instruction preset unit are not simply arranged side-by-side. (Source reconstruction unit output) The calibrated electrical angular velocity is used to characterize the rotational coupling relationship in the reference model, reducing the impact of encoder quantization noise and differential glitches on the reference flux reconstructing process from the angular velocity signal source side. The command preset unit generates the q-axis given current and provides the d-axis given current under vector control conditions where the d-axis given current is zero. The current loop PI controller in the flux domain MRAS identification unit outputs d-axis voltage commands and q-axis voltage commands respectively based on the errors between the d-axis and q-axis given currents and the corresponding d-axis and q-axis current feedback values, reducing the impact of dynamic current tracking error on the reference flux reconstructing process during speed changes from the current command source side. Thus, angular velocity noise and dynamic current tracking error are synchronously weakened before flux error is formed, reducing both high-frequency noise and transient bias in the error signals driving intermediate parameters b and c, thereby forming a synergistic suppression effect at the MRAS error signal source.

[0016] To achieve the above objective, the identification device performs the following identification method, including the following steps:

[0017] Step 1: Source reconstruction. Second-order closed-loop calibration is used to ensure consistency of the encoder's original angles, obtaining... Calibrate electrical angular velocity and Internal electrical angle ;

[0018] Step 2: Command preset, generate q-axis feedforward current based on mechanical dynamics equations, and superimpose it with the speed loop PI output to form q-axis given current;

[0019] Step 3: Magnetic flux linkage domain MRAS identification, construction of reference model and adjustable model, using the output of Step 1. The reference flux linkage is calculated online based on the d-axis and q-axis voltage commands output by the PI controller in the current loop of the MRAS identification unit, which calibrates the electric angular velocity and flux linkage domain. The adjustable flux linkage is calculated based on the inverse estimate of the inductance, the flux linkage-inductance ratio estimate, and the d-axis and q-axis current feedback values. The deviation between the reference flux linkage and the adjustable flux linkage is converted into a parameter correction value after decoupling and filtering.

[0020] Step 4: Gain scheduling. A cumulative correction value is generated based on the shaping drive signal. This cumulative correction value is superimposed on the initial intermediate parameter estimate to obtain the updated inverse inductance estimate and flux-inductance ratio estimate. The equivalent inductance and equivalent permanent magnet flux are then calculated. The current loop PI controller parameters are updated in real time through the gain scheduling strategy. The updated inverse inductance estimate and flux-inductance ratio estimate are fed back to the adjustable model for use in calculating the adjustable flux in the next control cycle.

[0021] The following provides a more detailed explanation of each step.

[0022] Step 1: Source reconstruction. Second-order closed-loop calibration is used to ensure consistency of the encoder's original angles, obtaining... Calibrate electrical angular velocity and Internal electrical angle ;

[0023] The control period is denoted as In the In the nth control cycle, the encoder outputs the nth... Original electrical angle for each control cycle The encoder output of the previous control cycle Original electrical angle for each control cycle To obtain a rough estimate of the electric angular velocity, first... and The difference is calculated to obtain the original electrical angle increment for adjacent control cycles. Since the electrical angle is a periodic quantity, it has... Periodicity; if subtracted directly, the result will be across 0 and... Angle jumps may occur at the boundary, so the original electrical angle increment needs to be wrapped before being divided by the control period. , obtained the Rough estimate of electric angular velocity per control cycle .

[0024] (1)

[0025] in, For the first The original electrical angle for each control cycle is expressed in radians (rad). For the first One control cycle of the original electrical angle; For the first A rough estimate of the electrical angular velocity for each control cycle is given, in radians per second (rad / s). Rough estimate of electric angular velocity per control cycle It has a fast response speed, but it is superimposed with encoder quantization noise and high-frequency glitches caused by mechanical jitter, making it unsuitable for direct use in subsequent modeling.

[0026] To obtain smooth and consistent rotational speed and angle signals, this invention employs a source reconstruction unit to perform closed-loop filtering on the noisy rotational speed and angle signals. The source reconstruction unit generates [the necessary parameters] through a closed-loop feedback mechanism. Internal electrical angle and The electrical angular velocity is calibrated. Its working principle is that in each control cycle, the source reconstruction unit reads the data saved at the end of the previous cycle. The internal electrical angle is compared with the original electrical angle measured by the encoder to obtain the phase error. After proportional-integral adjustment, the coarse estimate of the electrical angular velocity is corrected to obtain... Calibrate the electric angular velocity, then use this... The next control cycle is obtained by calibrating the electrical angular velocity. Internal electrical angle, for use in the next cycle.

[0027] In the The first control cycle first reads the data from memory from the previous cycle (the first control cycle). At the end of the control cycle, the result is obtained according to equation (7). Control cycle Internal electrical angle . No. Original electrical angle for each control cycle With the Control cycle Internal electrical angle By subtracting, the original angle deviation is obtained. Its expression is:

[0028] (2)

[0029] Since the electrical angle is a circumferential quantity, it has The periodicity. When subtracting directly, if and Located respectively On either side of the transition point, amplitudes will be close. To eliminate spurious large errors caused by periodic jumps, a wrapping function is introduced. This function takes any input angle value and adds or subtracts it. Multiples of integers, forced mapping to Within the main value range. After wrapping, the first value can be obtained. Phase error per control cycle :

[0030] (3)

[0031] in, For the first The phase error per control cycle, measured in radians (rad), is the driving signal for the source reconstruction unit to perform self-correction. When When, explain the first Control cycle Internal electrical angle Lagging behind Original electrical angle for each control cycle It needs to increase its internal rotation speed to catch up; when When, explain the first Control cycle Internal electrical angle ahead of the first Original electrical angle for each control cycle It is necessary to reduce the internal speed to wait.

[0032] Obtain the Phase error per control cycle After that, it cannot be used directly to correct the first... Control cycle Internal electrical angle Otherwise, encoder noise will enter unchanged. Calibrate the electric angular velocity. This invention employs a proportional-integral loop filter to process the phase error, generating a smooth speed correction term. (Continuous domain) Time and speed correction item The expression is:

[0033] (4)

[0034] in, For continuous domain Time and speed correction items For continuous domain Timing phase error, for Proportional gain for Integral gain The phase error from time zero to the current time The points.

[0035] In the One control cycle, let the sampling time be To facilitate implementation by a digital controller, the continuous-time signal is... and exist The sampled values ​​at each location are denoted as follows: and Therefore, the continuous integral term in equation (4) is approximated by discrete accumulation: To obtain the discrete form of the first Speed ​​correction item per cycle Its expression is:

[0036] (5)

[0037] in, For the first The rotational speed correction term for each cycle, in radians per second (rad / s). Accumulated phase error The summation index variable is used. The proportional channel provides an immediate response based on the current phase error, while the integral channel continuously accumulates historical errors through an accumulator, gradually eliminating steady-state error.

[0038] The first Rough estimate of electric angular velocity per control cycle With the Speed ​​correction item per cycle Add them together to get the first one. One cycle Calibrate electrical angular velocity :

[0039] (6)

[0040] in, For the first One cycle The electrical angular velocity is calibrated, in radians per second (rad / s). Rough estimate of electric angular velocity per control cycle Provides a fast dynamic skeleton in the speed signal to ensure rapid response during acceleration and deceleration; Speed ​​correction item per cycle Fine-tuning was performed to smooth out the burrs and static error attached to the dynamic frame. The two were then combined to obtain a speed signal that was both fast and smooth.

[0041] After obtaining the calibrated electrical angular velocity, the source reconstruction unit generates the first... Control cycle Internal electrical angle. During the control cycle. Under sufficiently short conditions, it can be considered It remains constant within this period, therefore the first Control cycle Internal electrical angle The formula is:

[0042] (7)

[0043] in, For the first Control cycle The internal electrical angle is the same as the first one in equation (3). Original electrical angle for each control cycle Compare to calculate the value of the phase error; For the first Control cycle Internal electrical angle, will be in the first Each cycle as a new Substitute the internal electrical angle into equation (3) and continue the closed-loop iteration.

[0044] At the moment of system power-on startup ( ), the 0th control cycle Internal electrical angle The initial value is set to If the system is equipped with an absolute encoder, the encoder angle read at the moment of power-on can also be directly assigned. This achieves phase alignment at the initial moment.

[0045] After processing by the aforementioned source reconstruction unit, the output of each cycle is Calibrate electrical angular velocity and Internal electrical angle It is a set of phase-consistent, dynamically smooth electrical angular velocities and electrical angles. This set of signals serves as the input signal for subsequent coordinate transformations and reference model calculations, replacing the original encoder signal in related calculations and suppressing signal disturbances caused by quantization noise and angle inconsistencies at the source.

[0046] Step 2: Command preset, generate q-axis feedforward current based on mechanical dynamics equations, and superimpose it with the speed loop PI output to form q-axis given current.

[0047] First, the definitions of the symbols required for this step are given: is the viscous damping coefficient, with units of Newton-meter-second per radian (N·m·s / rad). The moment of inertia is expressed in kilogram-meter (kg·m²). The load torque is expressed in Newton-meter (N·m). This is the torque coefficient, measured in Newton-meter per ampere (N·m / A). This represents the number of pole pairs of the motor. for Calibrate the electrical angular velocity; for The formula for calibrating mechanical angular velocity, expressed in radians per second (rad / s), is as follows: ; For the first One cycle Calibrate the electrical angular velocity; For the first Control cycle The formula for calibrating mechanical angular velocity, expressed in radians per second (rad / s), is as follows: ; This is the speed command, which is the target mechanical angular velocity that the motor is expected to achieve, in radians per second (rad / s). The q-axis current feedback value is measured by a current sensor and obtained through coordinate transformation, and the unit is ampere (A). This is the q-axis current reference component generated by the speed loop PI controller based on the speed deviation, in amperes (A). This is the q-axis feedforward current, measured in amperes (A).

[0048] The feedforward compensation method proposed in this invention operates under non-maximum torque-current ratio (non-MTPA) conditions. MTPA stands for maximum torque-current ratio, and its goal is to maximize the output torque by simultaneously adjusting the d-axis and q-axis currents given a total current. Non-MTPA conditions refer to operating states where the maximum torque-current ratio strategy is not employed. In this invention, a vector control method is used where the d-axis current is zero. Set to zero; under the closed-loop control of the current loop, the d-axis current feedback value It is usually close to zero, but not limited to being constantly equal to zero during dynamic processes. The reason why this invention chooses the non-MTPA operating condition is that when When set to zero, the electromagnetic torque is directly proportional to the q-axis current. This simple relationship makes analytical calculation of the q-axis feedforward current possible; simultaneously, in the parameter identification in the subsequent step three, Setting it to zero also allows the motor to be described using only a single equivalent inductance and equivalent permanent magnet flux linkage, avoiding the multi-parameter coupling problem caused by the different inductances of the d and q axes.

[0049] The mechanical equations of motion for the motor rotor are given by the rotational form of Newton's second law:

[0050] (8)

[0051] in, Electromagnetic torque is generated by the interaction between the motor current and the magnetic field, and its unit is Newton-meter (N·m).

[0052] In non-MTPA operating conditions ( Under these conditions, the electromagnetic torque is mainly contributed by the q-axis current, and the two are directly proportional:

[0053] (9)

[0054] Substituting equation (9) into equation (8), eliminate The following formula is obtained:

[0055] (10)

[0056] The core idea of ​​feedforward compensation is not to wait for the actual speed deviation to occur before the speed loop PI correction occurs, but to directly calculate the required q-axis feedforward current in advance based on the speed command and apply it to the current loop beforehand. In the analytical derivation of the q-axis feedforward current, the desired tracking state is used as the calculation assumption, that is, let... Calibrate mechanical angular velocity With speed command Consistency, that is And set the q-axis current feedback value q-axis feedforward current Provide, i.e. Substituting the above relationship into equation (10) and simplifying, we obtain the q-axis feedforward current. expression:

[0057] (11)

[0058] In order for the motor to keep up with the speed command, a q-axis feedforward current needs to be injected in advance. This current must overcome the inertial torque required for rotor acceleration. Viscous damping torque and external load torque These are the three parts of resistance.

[0059] In practical engineering, moment of inertia and viscous damping coefficient The nominal value can usually only be obtained through offline identification, and there is a deviation between it and the actual value; load torque Most of these parameters are unknown and time-varying. If equation (11) is used directly, inaccurate parameters will lead to an excessively large (overcompensated, causing overshoot) or an excessively small (undercompensated, resulting in insignificant effect) q-axis feedforward current. To solve this problem, this invention introduces an adjustable feedforward gain. Furthermore, the speed command in equation (11) and its derivative Essentially, it is a quantity that changes with time. To explicitly express the q-axis feedforward current, the time variables in equation (11) are explicitly labeled as time. The function, to obtain q-axis feedforward current at time 1 Function expression:

[0060] (12)

[0061] in, for Time-of-flight speed command; This is an adjustable feedforward gain, a dimensionless positive number. The adjustable feedforward gain is equivalent to a feedforward force knob; when... When full feedforward is applied, it indicates complete trust in the nominal mechanical parameters; when When the feedforward channel is closed, the system degenerates into pure feedback control; when When partial feedforward is applied, conservative compensation is used, which is suitable for situations with large parameter uncertainties; when When applying excess feedforward, it can be used to compensate for unmodeled additional drag or rotational inertia.

[0062] In practical digital controllers, directly... Performing differentiation operations can easily amplify quantization noise, causing unnecessary high-frequency jitter in the q-axis feedforward current. This is because the transfer function of pure differentiation operations in direct numerical differentiation (such as first-order backward difference) is... Its amplitude-frequency characteristic increases linearly with increasing frequency, significantly amplifying high-frequency components in the signal. However, the speed command signal inevitably contains slight quantization noise, which is severely amplified after pure differentiation, contaminating the q-axis feedforward current signal and subsequently injecting it into the motor through the current loop, causing unnecessary torque ripples and vibrations. To suppress this effect, this invention uses a bandwidth-limited differentiator instead of pure differentiation. In the Laplace domain, the transfer function of pure differentiation is... This invention uses a bandwidth-limited differentiator to replace pure differentiation operations, and the bandwidth-limited differentiator transfer function... The formula is:

[0063] (13)

[0064] in, For the Laplace operator; Here is the feedforward channel time constant, in seconds (s). The frequency response of this differentiator is: at frequencies much lower than... In the low-frequency band, the bandwidth-limited differentiator transfer function output is a pure differential, i.e. It can accurately reflect the rate of change of instructions; at the turning frequency In the nearby mid-frequency band, the bandwidth-limited differentiator's transfer function begins to transition from a differential characteristic to a constant gain, with increased phase lag, thus acting as a smoothing filter; at frequencies much higher than... High frequency band, molecules terms and denominators The terms cancel each other out, and the gain approaches the mean. The constant no longer increases with increasing frequency, thus effectively suppressing the amplification of high-frequency noise.

[0065] In equation (12), the bandwidth-limited differentiator transfer function is used. Algebraic differential operations are: And perform a Laplace transform on each term of equation (12), The Laplace transform is At the same time, taking into account and The term can be compensated for by adjusting the adjustable feedforward gain, ultimately yielding the frequency domain q-axis feedforward current. Function expression:

[0066] (14)

[0067] in, for Laplace transform of the time-speed command for Laplace transform of the q-axis feedforward current at time t.

[0068] To transform it into a difference equation that can be executed in a digital controller, intermediate variables are first introduced. Let it be: Equation (14) can then be written as:

[0069] (15)

[0070] For intermediate variables Cross-multiplication and rearranging the terms of the definition yields:

[0071] (16)

[0072] Convert it back to the time domain correspond This yields the first-order ordinary differential equation satisfied by the intermediate variables:

[0073] (17)

[0074] In the Each control cycle is discretized using the forward Euler method, and the sampling time is... Continuous variables and The sampled values ​​at that moment were respectively and In the first Each control cycle is represented by the forward difference instead of the differential. After substituting and rearranging, the solution is:

[0075] (18)

[0076] Through the discretization process described above, the continuous intermediate variables in the original continuous domain differential equation are... Transformed into discrete intermediate variables And thus, a difference equation that can be recursively computed in a digital signal processor was obtained. When the system powers on, and Initialization. The digital signal processor only needs to store the data from the previous cycle in each control cycle. Compared to the previous cycle Read the current Then, update recursively according to formula (18). .

[0077] In the frequency domain, the expression for the q-axis feedforward current is equation (15). In the time domain, Derivative of the intermediate variable , correspond Time and speed command , Corresponding constant Converting the above frequency domain expression back to the time domain yields: In the Each control cycle is represented by the forward difference instead of the differential. Continuous variables The sampled value at that moment After sorting, the first... q-axis feedforward current per control cycle expression:

[0078] (19)

[0079] In the The first control cycle, the... Speed ​​deviation per control cycle The expression is:

[0080] (20)

[0081] in, For the first Speed ​​command per control cycle For the first Control cycle Calibrate the mechanical angular velocity.

[0082] The digital signal processor maintains a speed loop integral accumulation variable in memory, denoted as the th . The value updated after each control cycle Its recursive calculation formula is:

[0083] (twenty one)

[0084] in, For the first The value updated after each control cycle is initialized to [value] when the system powers on. .

[0085] Finally, the speed loop PI controller outputs the q-axis current reference component based on the speed deviation. :

[0086] (twenty two)

[0087] in, This is the proportional gain of the speed loop. This is the integral gain of the speed loop.

[0088] The first q-axis feedforward current per control cycle The q-axis current reference component generated by the speed loop PI controller based on the speed deviation Superimposed, forming the final first The q-axis given current in each control cycle :

[0089] (twenty three)

[0090] The compensation effect of the q-axis feedforward current is mainly reflected in the frequency range of speed command changes. In steady-state and gradually changing processes, the speed command changes slowly, and the q-axis feedforward current provides torque compensation proportional to acceleration, reducing the dynamic burden on the speed loop PI controller. In the high-frequency range, due to the bandwidth-limited differentiator gain automatically rolling off, the q-axis feedforward current hardly responds to the high-frequency noise components in the command, preventing noise from being amplified and injected into the current loop through the feedforward channel. This frequency characteristic allows feedforward compensation to balance dynamic response speed and signal smoothness, providing sufficient compensation in the low-frequency range where fast response is required, and automatically attenuating in the high-frequency range where noise suppression is required.

[0091] Step 3: Magnetic flux linkage domain MRAS identification, construction of reference model and adjustable model, using the output of Step 1. The reference flux linkage is calculated online based on the d-axis and q-axis voltage commands output by the PI controller in the current loop of the MRAS identification unit, which calibrates the electric angular velocity and flux linkage domain. The adjustable flux linkage is calculated based on the inverse estimate of the inductance, the flux linkage-inductance ratio estimate, and the d-axis and q-axis current feedback values. The deviation between the reference flux linkage and the adjustable flux linkage is converted into a parameter correction value after decoupling and filtering.

[0092] The goal of this step is to identify the equivalent inductance of two physical parameters online. and equivalent permanent magnet flux .in, The equivalent inductance, measured in Henry (H), characterizes the ability of a winding to impede changes in current and determines how fast the current can change when a voltage is applied. The equivalent permanent magnet flux linkage, measured in Weber (Wb), characterizes the strength of the magnetic field of a permanent magnet and determines the torque generated per unit current.

[0093] In a permanent magnet synchronous motor, the equivalent inductance and equivalent permanent magnet flux Numerically, there are usually significant differences, with a difference of one or even several orders of magnitude. If directly compared in MRAS... and When two parameters are identified online, the corresponding error gradients and parameter update step sizes will be mismatched due to the severe asymmetry in the numerical scale of the two parameters. When adjusting the same adaptive gain, the parameter channel with the larger value will respond too aggressively and easily generate violent oscillations, while the parameter channel with the smaller value will respond slowly and remain almost motionless, causing the two parameters to fail to converge synchronously, making the system extremely difficult to tune.

[0094] To address this scale inconsistency issue, this invention mathematically reconstructs the parameters to be identified. Two new intermediate identification parameters are defined: the reciprocal of the inductance. Compared to magnetic flux inductance The formula is:

[0095] (twenty four)

[0096] in, It is the reciprocal of the inductance, and the unit is per henry (1 / H). This is the flux linkage inductance ratio, measured in amperes (A).

[0097] During the online identification process, the system maintains the intermediate parameter, the reciprocal of the inductance. Compared to magnetic flux inductance The estimated value is denoted as the inverse estimate of the inductance. Estimated value of flux linkage inductance In the first The system uses the first control cycle. Inductance reciprocal estimate for each control cycle and the Estimated flux inductance ratio for each control cycle Perform flux linkage calculation, and then update the result based on error feedback to obtain the first... Inductance reciprocal estimate for each control cycle and the Estimated flux inductance ratio for each control cycle When actual physical parameters are needed, the estimated equivalent inductance can be obtained through inverse calculation. and equivalent permanent magnet flux estimates :

[0098] (25)

[0099] in, The equivalent inductance is estimated by... The reciprocal is used to obtain the result, with the unit being Henry (H). The equivalent permanent magnet flux linkage is estimated by... Divide by The result is given in Weber (Wb).

[0100] Before constructing the reference model, we must first explain the two input quantities of the reference model: the d-axis voltage command. and q-axis voltage command The generation process. These two voltage commands are generated by the current loop PI controller, in the first... The first control cycle, for the d-axis, the first... The d-axis given current in each control cycle In the non-MTPA operating condition used in this invention, the value is set to zero. d-axis current feedback value per control cycle The current is measured by a current sensor and obtained through coordinate transformation. The d-axis current loop PI controller calculates the first... The d-axis given current in the first control cycle and the first control cycle The difference in the d-axis feedback current during the first control cycle is used to obtain the first... d-axis current error per control cycle :

[0101] (26)

[0102] The d-axis current loop PI controller performs proportional-integral calculations on this error and outputs the d-axis voltage command. : , This represents the d-axis current error; in a digital controller, the integral term is implemented using discrete accumulation, denoted as... For the first The cumulative value of the d-axis current error over each control cycle is derived using the following recursive formula: initial value Then the first D-axis voltage command for each control cycle for:

[0103] (27)

[0104] in, This is the d-axis current loop proportional gain. This is the integral gain of the d-axis current loop.

[0105] For the q-axis, the first The q-axis given current in each control cycle It consists of two superimposed parts. The first part is the q-axis current reference component generated by the speed loop PI controller based on the speed deviation. The second part is the q-axis feedforward current per control cycle The sum of the two is equation (23) in step two: Calculation of the q-axis current loop PI controller The q-axis given current in each control cycle With the q-axis current feedback value per control cycle The difference is used to obtain the first... q-axis current error per control cycle :

[0106] (28)

[0107] After proportional-integral calculation, the q-axis voltage command is output. : In digital controllers, The error is the q-axis current; the integral term is implemented using discrete accumulation, denoted as... For the first The cumulative value of the q-axis current error over each control cycle is derived using the following formula: initial value Then the first Q-axis voltage command per control cycle for:

[0108] (29)

[0109] in, This is the proportional gain of the q-axis current loop. This is the integral gain of the q-axis current loop.

[0110] The above d-axis current loop proportional gain d-axis current loop integral gain q-axis current loop proportional gain q-axis current loop integral gain In step four, the equivalent inductance and equivalent permanent magnet flux identified by MRAS will be updated online (see equations (50) and (51)).

[0111] Thus, in each control cycle, the current loop PI controller outputs the d-axis voltage command. and q-axis voltage command These two voltage commands, on the one hand, drive the inverter and control the motor operation in the vector control closed loop after inverse Park transformation and SVPWM modulation; on the other hand, they are also fed as inputs into the reference model below to participate in the online calculation of stator flux linkage.

[0112] Then, a reference model and an adjustable model are constructed respectively.

[0113] The reference model is based on the stator voltage equations and does not depend on any parameters to be identified, serving as a neutral identification benchmark. In the synchronously rotating dq coordinate system, the aforementioned d-axis voltage command is used. q-axis voltage command d-axis current feedback value q-axis current feedback value and the result obtained in step one Calibrate electrical angular velocity The stator flux linkage can be reconstructed online using the stator voltage equation:

[0114] (30)

[0115] in, This is the stator resistance, measured in ohms (Ω). The result obtained in step one Calibrate the electrical angular velocity, in radians per second (rad / s). d-axis reference flux linkage Here is the q-axis reference flux linkage, in Weber (Wb). The right-hand side of the equation... and The two terms are rotating back electromotive force terms, reflecting the cross-coupling caused by coordinate rotation. Additionally, The calibrated electrical angular velocity is used to characterize the rotational coupling relationship in the reference model, which can reduce the impact of encoder angular velocity noise on the reference flux reconstruction process; the d-axis voltage command and q-axis voltage command are respectively output by the current loop PI controller based on the error between the given d-axis and q-axis currents and the corresponding d-axis and q-axis current feedback values, and are correlated with... The calibrated electric angular velocity and d-axis and q-axis current feedback values ​​are used together to calculate the reference flux linkage of the reference model. The d-axis and q-axis voltage commands and current feedback values ​​characterize the impact of the current loop dynamics on the reference flux linkage reconstruction. Therefore, the source reconstruction unit and the command preset unit improve the reference flux linkage reconstruction conditions from the angular velocity signal source and the current command source, respectively, thereby reducing high-frequency noise and transient bias in the subsequent flux linkage error.

[0116] To implement this in a digital controller, in the first... In the first control cycle, the forward Euler method is used to discretize equation (30). Taking the d-axis as an example, in the first control cycle... One control cycle, in equation (30) respectively To replace the differential with a forward difference: Substituting into the d-axis voltage equation, we get:

[0117] (31)

[0118] Similarly, the discrete update formula for the q-axis reference flux can be obtained:

[0119] (32)

[0120] in, For the first Periodic d-axis reference flux and For the first Periodic q-axis reference flux; For the first Periodic d-axis reference flux and For the first The periodic q-axis reference flux is initially set to zero when the system is powered on.

[0121] The parameters used in the adjustable model are the inverse estimates of the inductance. Estimated value of flux linkage inductance If the estimation is accurate, the calculated results will be consistent with the reference model; if the estimation is inaccurate, there will be deviations. In non-MTPA operating conditions ( Under these conditions, the d-axis current is zero, and its secondary effects on magnetic circuit saturation and d- and q-axis cross-coupling can be ignored. Therefore, this invention uses a single equivalent inductor. and equivalent permanent magnet flux This describes the relationship between stator flux linkage and current. The original flux linkage equation is:

[0122] (33)

[0123] Substitute parameters and Substituting the values, we obtain the adjustable flux linkage along the d-axis. :

[0124] (34)

[0125] Similarly, the q-axis adjustable flux linkage is obtained. :

[0126] (35)

[0127] In the Each control cycle will transfer the actual parameters. , Replace with the first Inductance reciprocal estimate for each control cycle and the Estimated flux inductance ratio for each control cycle , and combined with the first d-axis current feedback value per control cycle and the q-axis current feedback value per control cycle The adjustable flux linkages along the d and q axes are obtained as follows:

[0128] (36)

[0129] in, For the first Adjustable d-axis flux for each cycle For the first The q-axis adjustable flux linkage for each cycle, in Weber (Wb). For the first d-axis current feedback value per control cycle No. q-axis current feedback value for each control cycle; For the first The inverse estimate of the inductance per control cycle. For the first Estimated flux linkage inductance ratio for each control cycle.

[0130] Therefore, within a control cycle, the system simultaneously possesses the reference flux provided by the reference model and the adjustable flux provided by the adjustable model. The difference between the two is the basis for updating the driving parameters, which will be elaborated in step four.

[0131] Step 4: Gain scheduling. A cumulative correction value is generated based on the shaping drive signal. This cumulative correction value is superimposed on the initial intermediate parameter estimate to obtain the updated inverse inductance estimate and flux-inductance ratio estimate. The equivalent inductance and equivalent permanent magnet flux are then calculated. The current loop PI controller parameters are updated in real time through the gain scheduling strategy. The updated inverse inductance estimate and flux-inductance ratio estimate are fed back to the adjustable model for use in calculating the adjustable flux in the next control cycle.

[0132] In the In each control cycle, the components of the reference flux and the adjustable flux on the d and q axes are subtracted to obtain the flux error:

[0133] (37)

[0134] in, For the first The d-axis flux error per control cycle For the first The q-axis flux linkage error per control cycle is expressed in Weber (Wb). These two errors collectively reflect... and The degree of deviation from the truth, when and When the value is zero, the identification is complete.

[0135] but and It simultaneously includes parameter deviation information and current operating condition information, requiring the construction of corresponding parameters based on parameter sensitivity relationships. and The instantaneous drive signal is constructed based on the design criteria of Popov's superstability theory. To ensure asymptotic stability in the parameter adaptive update process, the instantaneous drive signal should be constructed by combining flux linkage error and current feedback signals, and together with the subsequent linear forward channel and PI adaptive law, satisfy the strict positive reality design requirements. Based on the above principles, the instantaneous drive signal is constructed as follows:

[0136] (38)

[0137] in, For the first Driven by one control cycle The instantaneous drive signal integrates the flux linkage error of the d and q axes as well as current information; its amplitude and sign indicate... The direction and magnitude of the deviation; For the first Driven by one control cycle The instantaneous drive signal is directly used with respect to the d-axis flux linkage error, because The change in flux mainly affects the amplitude of the d-axis flux linkage. Under vector control conditions where the d-axis current is zero, Typically close to zero, the two instantaneous drive signal channels have good decoupling characteristics.

[0138] and The parameter estimate is obtained by multiplying and adding the flux linkage error and the instantaneous current value, inevitably carrying current sampling noise and encoder residual ripple. If these instantaneous drive signals are directly fed into the adaptive law, the parameter estimate will fluctuate or even oscillate at high frequency. Therefore, this invention inserts a first-order low-pass forward channel between the error signal and the adaptive update law to perform dynamic shaping.

[0139] In the continuous domain, the differential equation of a first-order low-pass filter is:

[0140] (39)

[0141] (40)

[0142] in, for The forward channel gain of a channel controls the filtering bandwidth of that channel and determines... For instantaneous error signals The balance between tracking speed and high-frequency suppression capability; for The forward channel gain of a channel controls the filtering bandwidth of that channel and determines... For instantaneous error signals The balance between tracking speed and high-frequency suppression capability.

[0143] To implement this in a digital controller, forward Euler method is used to discretize equations (39) and (40). For example, a channel:

[0144] The differential can be replaced by the forward difference: Substitution formula (39):

[0145] (41)

[0146] Final result:

[0147] (42)

[0148] Similarly, we can obtain The recursive formula for the channel:

[0149] (43)

[0150] in, For the first After one cycle of plastic surgery Shaping drive signal, For the first After one cycle of plastic surgery Shaping drive signal. Equations (42) and (43) mean that in each control cycle, the new output value is equal to the decay of the retained value from the previous cycle plus the weighted sum of the new input value for the current cycle. Gain , The larger the gain, the faster the attenuation, the larger the channel bandwidth, and the more responsive the response, but the weaker the noise suppression; the smaller the gain, the better. By selecting the appropriate gain... , , making , It can quickly track error trends without being swayed by individual sampling spikes.

[0151] Obtained shaping drive signal and Then, they need to be converted to pairs. and The cumulative correction amount. This invention employs a proportional-integral adaptive update law to obtain the cumulative correction amount over the continuous domain. and expression:

[0152] (44)

[0153] (45)

[0154] in, for Channel proportional gain, for Channel proportional gain, for Channel integral gain, for Channel integral gain.

[0155] In the For each control cycle, the proportional terms in equations (44) and (45) are directly taken as their current values, and the integral terms are replaced by discrete summation: , , obtain the discrete domain up to the th Cumulative correction amount per control cycle and expression:

[0156] (46)

[0157] (47)

[0158] in As of the date The cumulative correction to the inverse estimate of the inductance over each control cycle. As of the date The cumulative correction to the estimated flux-inductance ratio over each control cycle. For summation index variables.

[0159] The role of the PI structure here is that in the early stage of identification, when the parameter deviation is large and the driving signal amplitude is large, the proportional term provides a large step correction to quickly reduce the deviation; as the parameters approach the true value, the driving signal gradually decreases, the contribution of the proportional term shrinks, and at this time the integral term has accumulated enough historical information to continuously output a stable and gentle correction force, which gradually eliminates the residual small steady-state error, which is conducive to reducing the steady-state parameter deviation.

[0160] Finally, the number was obtained The cumulative correction for each control cycle is then added to the initial estimate to obtain the parameter estimate for the next control cycle.

[0161] (48)

[0162] in, For the first Inductance reciprocal estimate for each control cycle For the first The estimated flux linkage-to-inductance ratio for the first control cycle will be... The new current value of each control cycle is substituted into the adjustable model equation (36) to perform a new round of flux linkage calculation, forming a closed-loop iteration. Its update mechanism can be summarized as follows: In each control cycle, the system calculates the adjustable flux linkage based on the current parameter estimate and compares it with the reference flux linkage to obtain the flux linkage error; after dynamic shaping and proportional-integral adaptive update law, the flux linkage error forms the cumulative correction up to the current control cycle. The system uses the initial estimate as a benchmark and adds this cumulative correction to obtain the parameter estimate for the next control cycle, and substitutes it into the adjustable model equation (36) as the new current value to perform a new round of flux linkage calculation, forming a closed-loop iteration. As the iteration progresses, the adjustable flux linkage and the reference flux linkage tend to be consistent, and the parameter estimate gradually approaches the actual parameters. In the first control cycle... After the intermediate parameters are updated in each control cycle, the inverse calculation relationship defined in step three (25) is used to... , Convert to , The calculation formulas are as follows: , .

[0163] The identified parameters are used to update the control loop parameters in real time. This invention employs a gain scheduling strategy to map the identification results of the MRAS to the parameters of the current loop PI controller, achieving online tuning. First, based on the identified equivalent permanent magnet flux linkage... The target bandwidth of the current loop is calculated using the amplitude. :

[0164] (49)

[0165] in, The target bandwidth of the current loop is expressed in radians per second (rad / s). The base bandwidth is the lowest bandwidth when the flux linkage is zero. This represents the flux linkage modulation coefficient. When the identified flux linkage is strong, the target bandwidth is increased accordingly to fully utilize the motor's torque output capability.

[0166] Then, in the... Each control cycle is based on the target bandwidth and the identified equivalent inductance estimate. Update the current loop PI controller parameters according to the standard first-order current model design principles. For the d-axis current loop:

[0167] (50)

[0168] For the q-axis current loop:

[0169] (51)

[0170] in, This is the d-axis current loop proportional gain. This is the d-axis current loop integral gain; This is the proportional gain of the q-axis current loop. This is the integral gain of the q-axis current loop; This is the stator resistance.

[0171] This invention organically integrates signal purification by the source reconstruction unit, q-axis feedforward current compensation generated by the instruction preset unit, and adaptive identification by parameter reconstruction. This suppresses noise interference from the signal source, effectively suppresses identification errors under dynamic operating conditions, improves the online identification accuracy of equivalent inductance and equivalent permanent magnet flux linkage, and overcomes the control performance degradation problem caused by parameter mismatch through the integrated closed loop of identification and control.

[0172] The beneficial effects of this invention are:

[0173] 1. By using parametric reconstruction, the scale mismatch problem of multi-parameter identification is effectively solved. This invention directly identifies the equivalent inductance in traditional MRAS. and equivalent permanent magnet flux Reconstructed into the reciprocal of the inductance Compared to magnetic flux inductance Two intermediate parameters. This mathematical transformation reduces the numerical difference between the two parameters from nearly a hundred times to within one order of magnitude, allowing the two parameter channels in the adaptive law to be tuned with similar gains. This helps reduce tuning difficulties caused by inconsistent parameter scales, ensures a more balanced and coordinated convergence process for the two parameters, and reduces the complexity of engineering debugging.

[0174] 2. By introducing pre-set instructions, the adverse effects of dynamic operating conditions on identification accuracy are effectively suppressed. This invention introduces a q-axis feedforward current based on the mechanical dynamics equations into the current loop, providing the q-axis feedforward current in advance during motor acceleration and deceleration, thus reducing dynamic current tracking errors. This reduces the impact of dynamic current errors on flux linkage calculation in the transient process, ensuring the accuracy of the MRAS error signal under dynamic operating conditions from the source, and improving the convergence speed and smoothness of parameter identification during speed changes.

[0175] 3. By introducing a source reconstruction calibration mechanism, the stability of identification is improved from the signal source. This invention uses a source reconstruction unit to perform closed-loop phase error adjustment on the original encoder angle, effectively filtering out high-frequency disturbances introduced by encoder quantization noise and inconsistent angle measurements, thus obtaining smooth and consistent electrical angular velocity and electrical angle signals. The signal obtained after source reconstruction... Calibrate electric angular velocity and The internal electrical angles are used for the calculation of rotational coupling terms and coordinate transformation in the reference model, thereby improving the signal-to-noise ratio of the MRAS error signal, suppressing transient spikes and steady-state oscillations in the parameter identification results, and enhancing the robustness of the identification process under the non-MTPA and variable speed conditions.

[0176] 4. By integrating identification and control into a closed loop, online self-tuning of controller parameters is achieved. This invention employs a gain scheduling strategy, mapping the equivalent inductance and equivalent permanent magnet flux identified by MRAS to the proportional gain and integral gain of the current loop PI controller in real time. This allows the dynamic characteristics of the current controller to be adjusted online according to changes in motor parameters. This closed-loop mechanism avoids control performance degradation caused by factors such as temperature changes and magnetic circuit saturation, achieving collaborative optimization between the identification system and the control system, and improving the overall performance and reliability of the motor drive system under complex operating conditions.

[0177] 5. The source reconstruction unit of this invention is not used alone for speed feedback smoothing, nor is the command preset unit used alone for improving rotational speed response. Rather, both work together in the construction process of the flux linkage domain MRAS reference model and error signal. The source reconstruction unit reduces the contamination of coordinate transformation and rotational back EMF terms by angle and angular velocity noise, while the command preset unit reduces the dynamic deviation between the actual current and the given current during rotational speed changes. Together, they reduce both high-frequency noise and transient bias in the error signals of the driving intermediate parameters b and c. Therefore, this invention produces a synergistic suppression effect on the source of the MRAS error signal, rather than a simple parallel combination of angle filtering, q-axis feedforward compensation, and MRAS identification. Attached Figure Description

[0178] Figure 1 This is a structural block diagram of the device of the present invention.

[0179] Figure 2 This is a flowchart of the method of the present invention.

[0180] Figure 3 This is a flowchart of the speed and angle consistency calibration based on source reconstruction.

[0181] Figure 4 This is a flowchart for generating the q-axis feedforward current based on a preset instruction.

[0182] Figure 5 Flowcharts for constructing the reference and adjustable models of the magnetic flux domain MRAS.

[0183] Figure 6 A flowchart for constructing the instantaneous error signal and the dynamic shaping of the linear forward channel is provided.

[0184] Figure 7 A comparison of the speed transient response under different adjustable feedforward gains;

[0185] Figure 8 A comparison chart of rotational speed response and current with and without signal source reconstruction.

[0186] Figure 9 The waveform of the back electromotive force of the experimental motor is shown.

[0187] Figure 10 The figure shows the online estimation results of parameters b and c under different MRAS identification structures.

[0188] Figure 11 The waveforms of the d-axis and q-axis current response of the motor under step conditions without source reconstruction and command preset are shown.

[0189] Figure 12 The speed response waveform of the motor under step conditions without source reconstruction and command preset is shown.

[0190] Figure 13 The d-axis and q-axis current response waveforms of the motor under step conditions with added source reconstruction and command preset.

[0191] Figure 14 The motor speed response waveform under step conditions is obtained by incorporating source reconstruction and command preset.

[0192] Figure 15 The waveforms of the d-axis and q-axis current response of the motor under sinusoidal operating conditions without source reconstruction and command preset are shown.

[0193] Figure 16 The speed response waveform of the motor under sinusoidal operating conditions without source reconstruction and command preset is shown.

[0194] Figure 17 The current response waveforms of the motor under sinusoidal operating conditions with added source reconstruction and command preset are shown.

[0195] Figure 18 The motor speed response waveform under sinusoidal operating conditions is incorporated with source reconstruction and command preset. Detailed Implementation

[0196] See Figures 1 to 2 The permanent magnet synchronous motor parameter identification device disclosed in this invention includes a source reconstruction unit, an instruction preset unit, a flux linkage domain MRAS identification unit, and a gain scheduling unit. Each unit can be implemented in a digital signal processor through programming.

[0197] Reference Figure 1 This invention provides a permanent magnet synchronous motor (PMSM) identification device that coordinates source reconstruction and command presetting. The device includes a source reconstruction unit, a command presetting unit, a flux linkage domain MRAS identification unit, and a gain scheduling unit. The source reconstruction unit includes a speed predictor, a phase detector with wrapping processing, a PI loop filter, and an integrator. The command presetting unit includes electromagnetic torque mapping, mechanical dynamics feedforward compensation, and a speed loop PI controller. The flux linkage domain MRAS identification unit includes a current loop PI controller, a reference model, and an adjustable model. The gain scheduling unit includes an error signal construction module, a linear forward channel, a PI adaptive law, and a parameter-to-PI controller parameter mapping module. The identification device works in conjunction with a coordinate transformation module, an SVPWM modulation module, an inverter, a DC power supply, the PMSM, and an encoder to achieve PMSM parameter identification and online updating of current loop control parameters. Figure 1 No superscript in the middle The variable represents the general form of the corresponding signal; in the derivation of discrete control, its first... The value of each control cycle is above the standard. express.

[0198] The encoder is used to detect the rotor position information of a permanent magnet synchronous motor (PMSM) and output the raw electrical angle. To the source reconstruction unit. The velocity predictor is based on the original electrical angle. A rough estimate of the electric angular velocity was obtained. The phase detector with encapsulation is based on the original electrical angle. and Internal electrical angle Obtain the phase error PI loop filter based on phase error Output speed correction item and a rough estimate of the electric angular velocity. Superimposed Calibrate electrical angular velocity . Calibrate electrical angular velocity Obtained through the integration process Internal electrical angle And converted by the pole logarithm conversion module Calibrate mechanical angular velocity .

[0199] The instruction preset unit receives the speed instruction. , Calibrate mechanical angular velocity and q-axis current feedback value Speed ​​command and Calibrate mechanical angular velocity The difference is used to obtain the speed error. The speed loop PI controller is based on the speed error. Output q-axis current reference component Electromagnetic torque mapping and mechanical dynamics feedforward compensation generate q-axis feedforward current based on speed commands and motor mechanical parameters. The q-axis feedforward current With q-axis current reference component Superimposed to form the given current along the q-axis Under vector control conditions where the d-axis current is zero, the d-axis current is... Set to zero. The given current along the d-axis. and q-axis given current The data is input to the current loop PI controller within the magnetic flux domain MRAS identification unit.

[0200] The current loop PI controller within the flux linkage domain MRAS identification unit receives the d-axis given current. q-axis given current d-axis current feedback value and q-axis current feedback value Wherein, the d-axis given current. With d-axis current feedback value The difference is used to obtain the d-axis current error. q-axis given current With q-axis current feedback value The difference is used to obtain the q-axis current error. The current loop PI controller is based on the d-axis current error. and q-axis current error Output d-axis voltage command and q-axis voltage command .

[0201] Magnetic flux domain MRAS identification unit utilizes Calibrate electrical angular velocity d-axis voltage command q-axis voltage command d-axis current feedback value and q-axis current feedback value Construct a reference model to obtain the reference magnet link. Simultaneously, the magnetic flux domain MRAS identification unit estimates the inverse of the inductance value. Estimated value of flux linkage inductance ratio d-axis current feedback value and q-axis current feedback value Construct an adjustable model to obtain an adjustable magnetic flux. Reference magnet link and adjustable magnetic flux The input is fed to the gain scheduling unit, which calculates the difference to obtain the flux linkage error. The error signal construction module then generates corresponding inverse inductance estimates. Estimated value of flux linkage inductance Instantaneous drive signal and Instantaneous drive signal and After dynamic shaping via the linear forward channel, the shaping drive signal is obtained. and The inductance inverse estimate is updated by the PI adaptive law. Estimated value of flux linkage inductance .

[0202] The gain scheduling unit is based on the updated inverse inductance estimate. Estimated value of flux linkage inductance The parameters of the current loop PI controller are obtained by mapping the parameters to the PI controller parameters. and The updated current loop PI controller parameters are fed back to the current loop PI controller within the flux linkage domain MRAS identification unit; simultaneously, the updated inductance reciprocal estimate is also fed back. Estimated value of flux linkage inductance The adjustable model fed back to the MRAS identification unit in the flux linkage domain is used for adjustable flux linkage calculation in the next control cycle. The d-axis voltage command output by the current loop PI controller... and q-axis voltage command After coordinate transformation, the signal is sent to the SVPWM modulation module, which controls the inverter to drive the permanent magnet synchronous motor (PMSM). The encoder, current feedback, flux linkage domain MRAS identification unit, and gain scheduling unit together form a closed-loop identification and control process.

[0203] like Figure 2 As shown, the method flow of this invention includes source reconstruction, instruction preset, magnetic flux domain MRAS identification, and gain scheduling. The source reconstruction unit receives the raw electrical angle output by the encoder, performs closed-loop calibration on it, and outputs... Calibrate electric angular velocity and Internal electrical angle. The command preset unit receives speed commands. The mechanical angular velocity is calibrated, and after speed error calculation, speed loop PI control, electromagnetic torque mapping, and mechanical dynamics feedforward compensation, the q-axis setpoint current is output; under vector control conditions where the d-axis setpoint current is zero, the output d-axis setpoint current is zero. The flux linkage domain MRAS identification unit receives... The electric angular velocity, d-axis given current, q-axis given current, and d- and q-axis current feedback values ​​are calibrated. These are then processed by a current-loop PI controller to generate d- and q-axis voltage commands. Reference and adjustable models are constructed to obtain reference and adjustable flux linkages. The deviation between the reference and adjustable flux linkages is processed by an error signal construction module to generate a drive signal for parameter updates. The gain scheduling unit receives this drive signal, processes it through a linear forward channel and a PI adaptive law, and updates the inductance reciprocal estimate and flux linkage-to-inductance ratio estimate. The updated identification results are mapped to current-loop PI controller parameters and fed back to the current-loop PI controller within the flux linkage domain MRAS identification unit, achieving coordinated updates of parameter identification and control parameters.

[0204] In one specific embodiment, a surface-mounted permanent magnet synchronous motor was used as the controlled object for experimental verification. The main parameters of the experimental motor and control system are shown in Table 1. The parameters listed in Table 1 are only used to illustrate the experimental platform and control system initialization method of this embodiment and do not constitute a limitation on the applicable motor type, power rating, rated speed, or voltage rating of this invention. Table 1 gives the main rated parameters of the experimental motor, including rated frequency, rated power, rated speed, rated voltage, and stator resistance. Control cycle Equivalent inductance .in, and As the initial estimate of MRAS, the basic process of parameter updating remains unchanged under the condition of stable convergence of the adaptive law. It should be kept in line with the driver control cycle to ensure system timing synchronization.

[0205] Table 1 Main parameters of the experimental motor and control system

[0206] Motor parameters numerical values Rated frequency (Hz) 166.7 Rated power (kW) 1.8 Rated speed (rpm) 2500 Rated voltage (V) 72 Stator resistance (Ω) 0.053 Control cycle (s) 0.000125 Equivalent inductance (H) 0.000168

[0207] To achieve the above objectives, the present invention includes the following steps:

[0208] Step 1: Source reconstruction, such as Figure 3 The encoder's original angle is consistent using second-order closed-loop calibration to obtain... Calibrate electrical angular velocity and Internal electrical angle .

[0209] The control period is denoted as In the In the nth control cycle, the encoder outputs the nth... Original electrical angle for each control cycle The encoder output of the previous control cycle Original electrical angle for each control cycle To obtain a rough estimate of the electric angular velocity, first... and The difference is calculated to obtain the original electrical angle increment for adjacent control cycles. Since the electrical angle is a periodic quantity, it has... Periodicity; if subtracted directly, the result will be across 0 and... Angle jumps may occur at the boundary, so the original electrical angle increment needs to be wrapped before being divided by the control period. , obtained the Rough estimate of electric angular velocity per control cycle .

[0210] (1)

[0211] in, For the first The original electrical angle for each control cycle is expressed in radians (rad). For the first One control cycle of the original electrical angle; For the first A rough estimate of the electrical angular velocity for each control cycle is given, in radians per second (rad / s). Rough estimate of electric angular velocity per control cycle It has a fast response speed, but it is superimposed with encoder quantization noise and high-frequency glitches caused by mechanical jitter, making it unsuitable for direct use in subsequent modeling.

[0212] To obtain smooth and consistent rotational speed and angle signals, this invention employs a source reconstruction unit to perform closed-loop filtering on the noisy rotational speed and angle signals. The source reconstruction unit generates [the necessary parameters] through a closed-loop feedback mechanism. Internal electrical angle and The electrical angular velocity is calibrated. Its working principle is that in each control cycle, the source reconstruction unit reads the data saved at the end of the previous cycle. The internal electrical angle is compared with the original electrical angle measured by the encoder to obtain the phase error. After proportional-integral adjustment, the coarse estimate of the electrical angular velocity is corrected to obtain... Calibrate the electric angular velocity, then use this... The next control cycle is obtained by calibrating the electrical angular velocity. Internal electrical angle, for use in the next cycle.

[0213] In the The first control cycle first reads the data from memory from the previous cycle (the first control cycle). At the end of the control cycle, the result is obtained according to equation (7). Control cycle Internal electrical angle . No. Original electrical angle for each control cycle With the Control cycle Internal electrical angle By subtracting, the original angle deviation is obtained. Its expression is:

[0214] (2)

[0215] Since the electrical angle is a circumferential quantity, it has The periodicity. When subtracting directly, if and Located respectively On either side of the transition point, amplitudes will be close. The false large error. For example, near and When approaching zero, the two values ​​actually differ by only a very small angle, but a direct subtraction yields a value close to zero. The large deviation. To eliminate this spurious large error caused by periodic jumps, a wrapping function is introduced. This function takes any input angle value and adds or subtracts it. Multiples of integers, forced mapping to Within the main value range. After wrapping, the first value can be obtained. Phase error per control cycle :

[0216] (3)

[0217] in, For the first The phase error per control cycle, measured in radians (rad), is the driving signal for the source reconstruction unit to perform self-correction. When When, explain the first Control cycle Internal electrical angle Lagging behind Original electrical angle for each control cycle It needs to increase its internal rotation speed to catch up; when When, explain the first Control cycle Internal electrical angle ahead of the first Original electrical angle for each control cycle It is necessary to reduce the internal speed to wait.

[0218] Obtain the Phase error per control cycle After that, it cannot be used directly to correct the first... Control cycle Internal electrical angle Otherwise, encoder noise will enter unchanged. Calibrate the electric angular velocity. This invention employs a proportional-integral loop filter to process the phase error, generating a smooth speed correction term. (Continuous domain) Time and speed correction item The expression is:

[0219] (4)

[0220] in, For continuous domain Time and speed correction items On continuous domain Timing phase error, for Proportional gain for Integral gain The phase error from time zero to the current time The points.

[0221] In the One control cycle, let the sampling time be To facilitate implementation by a digital controller, the continuous-time signal is... and exist The sampled values ​​at each location are denoted as follows: and Therefore, the continuous integral term in equation (4) is approximated by discrete accumulation: To obtain the discrete form of the first Speed ​​correction item per cycle Its expression is:

[0222] (5)

[0223] in, For the first The rotational speed correction term for each cycle, in radians per second (rad / s). Accumulated phase error The summation index variable is used. The proportional channel provides an immediate response based on the current phase error, while the integral channel continuously accumulates historical errors through an accumulator, gradually eliminating steady-state error.

[0224] The first Rough estimate of electric angular velocity per control cycle With the Speed ​​correction item per cycle Add them together to get the first one. One cycle Calibrate electrical angular velocity :

[0225] (6)

[0226] in, For the first One cycle The electrical angular velocity is calibrated, in radians per second (rad / s). Rough estimate of electric angular velocity per control cycle Provides a fast dynamic skeleton in the speed signal to ensure rapid response during acceleration and deceleration; Speed ​​correction item per cycle Fine-tuning was performed to smooth out the burrs and static error attached to the dynamic frame. The two were then combined to obtain a speed signal that was both fast and smooth.

[0227] After obtaining the calibrated electrical angular velocity, the source reconstruction unit generates the first... Control cycle Internal electrical angle. During the control cycle. Under sufficiently short conditions, it can be considered It remains constant within this period, therefore the first Control cycle Internal electrical angle The formula is:

[0228] (7)

[0229] in, For the first Control cycle The internal electrical angle is the same as the first one in equation (3). Original electrical angle for each control cycle Compare to calculate the value of the phase error; For the first Control cycle Internal electrical angle, will be in the first Each cycle as a new Substitute the internal electrical angle into equation (3) and continue the closed-loop iteration.

[0230] At the moment of system power-on startup ( ), the 0th control cycle Internal electrical angle The initial value is set to If the system is equipped with an absolute encoder, the encoder angle read at the moment of power-on can also be directly assigned. This achieves phase alignment at the initial moment.

[0231] After processing by the aforementioned source reconstruction unit, the output of each cycle is Calibrate electrical angular velocity and Internal electrical angle It is a set of phase-consistent, dynamically smooth electrical angular velocities and electrical angles. This set of signals serves as the input signal for subsequent coordinate transformations and reference model calculations, replacing the original encoder signal in related calculations and suppressing signal disturbances caused by quantization noise and angle inconsistencies at the source.

[0232] Step 2: Preset instructions, such as Figure 4 The q-axis feedforward current is generated based on the mechanical dynamics equations and superimposed with the speed loop PI output to form the q-axis given current.

[0233] First, the definitions of the symbols required for this step are given: is the viscous damping coefficient, with units of Newton-meter-second per radian (N·m·s / rad). The moment of inertia is expressed in kilogram-meter (kg·m²). The load torque is expressed in Newton-meter (N·m). This is the torque coefficient, measured in Newton-meter per ampere (N·m / A). This represents the number of pole pairs of the motor. for Calibrate the electrical angular velocity; for The formula for calibrating mechanical angular velocity, expressed in radians per second (rad / s), is as follows: ; For the first One cycle Calibrate the electrical angular velocity; For the first Control cycle The formula for calibrating mechanical angular velocity, expressed in radians per second (rad / s), is as follows: ; This is the speed command, which is the target mechanical angular velocity that the motor is expected to achieve, in radians per second (rad / s). The q-axis current feedback value is measured by a current sensor and obtained through coordinate transformation, and the unit is ampere (A). This is the q-axis current reference component generated by the speed loop PI controller based on the speed deviation, in amperes (A). This is the q-axis feedforward current, measured in amperes (A).

[0234] The feedforward compensation method proposed in this invention operates under non-maximum torque-current ratio (non-MTPA) conditions. MTPA stands for maximum torque-current ratio, and its goal is to maximize the output torque by simultaneously adjusting the d-axis and q-axis currents given a total current. Non-MTPA conditions refer to operating states where the maximum torque-current ratio strategy is not employed. In this invention, a vector control method is used where the d-axis current is zero. Set to zero; under the closed-loop control of the current loop, the d-axis current feedback value It is usually close to zero, but not limited to being constantly equal to zero during dynamic processes. The reason why this invention chooses the non-MTPA operating condition is that when When set to zero, the electromagnetic torque is directly proportional to the q-axis current. This simple relationship makes analytical calculation of the q-axis feedforward current possible; simultaneously, in the parameter identification in the subsequent step three, Setting it to zero also allows the motor to be described using only a single equivalent inductance and equivalent permanent magnet flux linkage, avoiding the multi-parameter coupling problem caused by the different inductances of the d and q axes.

[0235] The mechanical equations of motion for the motor rotor are given by the rotational form of Newton's second law:

[0236] (8)

[0237] in, Electromagnetic torque is generated by the interaction between the motor current and the magnetic field, and its unit is Newton-meter (N·m).

[0238] In non-MTPA operating conditions ( Under these conditions, the electromagnetic torque is mainly contributed by the q-axis current, and the two are directly proportional:

[0239] (9)

[0240] Substituting equation (9) into equation (8), eliminate The following formula is obtained:

[0241] (10)

[0242] The core idea of ​​feedforward compensation is not to wait for the actual speed deviation to occur before the speed loop PI correction occurs, but to directly calculate the required q-axis feedforward current in advance based on the speed command and apply it to the current loop beforehand. In the analytical derivation of the q-axis feedforward current, the desired tracking state is used as the calculation assumption, that is, let... Calibrate mechanical angular velocity With speed command Consistency, that is And set the q-axis current feedback value q-axis feedforward current Provide, i.e. Substituting the above relationship into equation (10) and simplifying, we obtain the q-axis feedforward current. expression:

[0243] (11)

[0244] In order for the motor to keep up with the speed command, a q-axis feedforward current needs to be injected in advance. This current must overcome the inertial torque required for rotor acceleration. Viscous damping torque and external load torque These are the three parts of resistance.

[0245] In practical engineering, moment of inertia and viscous damping coefficient The nominal value can usually only be obtained through offline identification, and there is a deviation between it and the actual value; load torque Most of these parameters are unknown and time-varying. If equation (11) is used directly, inaccurate parameters will lead to an excessively large (overcompensated, causing overshoot) or an excessively small (undercompensated, resulting in insignificant effect) q-axis feedforward current. To solve this problem, this invention introduces an adjustable feedforward gain. Furthermore, the speed command in equation (11) and its derivative Essentially, it is a quantity that changes with time. To explicitly express the q-axis feedforward current, the time variables in equation (11) are explicitly labeled as time. The function, to obtain q-axis feedforward current at time 1 Function expression:

[0246] (12)

[0247] in, for Time-of-flight speed command; This is an adjustable feedforward gain, a dimensionless positive number. The adjustable feedforward gain is equivalent to a feedforward force knob; when... When full feedforward is applied, it indicates complete trust in the nominal mechanical parameters; when When the feedforward channel is closed, the system degenerates into pure feedback control; when When partial feedforward is applied, conservative compensation is used, which is suitable for situations with large parameter uncertainties; when When applying excess feedforward, it can be used to compensate for unmodeled additional drag or rotational inertia.

[0248] Suppose that the actual moment of inertia of a certain motor is slightly larger than its nominal value. If it is set according to the nominal value... The q-axis feedforward current is actually too low, and the speed will still lag behind the command briefly during acceleration. The technician can gradually increase the adjustable feedforward gain from 1.0, first trying 1.1 to observe the speed tracking; if the tracking error decreases but lag still exists, continue increasing to 1.2; if significant overshoot occurs when adjusting to 1.3, it indicates overcompensation, and the adjustable feedforward gain should be adjusted back to around 1.2. Through several trials, the optimal adjustable feedforward gain value can be found. This process only requires adjusting... This scalar parameter does not require separate precise calibration. , and This greatly reduces the complexity of engineering tuning.

[0249] Figure 7 For different adjustable feedforward gains A comparison of the transient response of the motor speed. (From...) Figure 7 It can be seen that without feedforward compensation, i.e. When the speed feedback increases slowly, it takes a long time to reach the target speed; when As the value is gradually increased from 0.05 and 0.10 to 0.15, the upward slope of the speed feedback curve increases, and the time for the system to reach the target speed gradually shortens, indicating that the command preset can improve the dynamic response capability during speed command changes. Because the speed dynamic lag is reduced, the dynamic error caused by the current command lag during speed changes is correspondingly reduced, which helps to reduce transient disturbances in the flux linkage domain MRAS reference model and provides more stable dynamic input conditions for subsequent parameter identification.

[0250] In practical digital controllers, directly... Performing differentiation operations can easily amplify quantization noise, causing unnecessary high-frequency jitter in the q-axis feedforward current. This is because the transfer function of pure differentiation operations in direct numerical differentiation (such as first-order backward difference) is... Its amplitude-frequency characteristic increases linearly with increasing frequency, significantly amplifying high-frequency components in the signal. However, the speed command signal inevitably contains slight quantization noise, which is severely amplified after pure differentiation, contaminating the q-axis feedforward current signal and subsequently injecting it into the motor through the current loop, causing unnecessary torque ripples and vibrations. To suppress this effect, this invention uses a bandwidth-limited differentiator instead of pure differentiation. In the Laplace domain, the transfer function of pure differentiation is... This invention uses a bandwidth-limited differentiator to replace pure differentiation operations, and the bandwidth-limited differentiator transfer function... The formula is:

[0251] (13)

[0252] in, For the Laplace operator; Here is the feedforward channel time constant, in seconds (s). The frequency response of this differentiator is: at frequencies much lower than... In the low-frequency band, the bandwidth-limited differentiator transfer function output is a pure differential, i.e. It can accurately reflect the rate of change of instructions; at the turning frequency In the nearby mid-frequency band, the bandwidth-limited differentiator's transfer function begins to transition from a differential characteristic to a constant gain, with increased phase lag, thus acting as a smoothing filter; at frequencies much higher than... High frequency band, molecules terms and denominators The terms cancel each other out, and the gain approaches the mean. The constant no longer increases with increasing frequency, thus effectively suppressing the amplification of high-frequency noise.

[0253] In equation (12), the bandwidth-limited differentiator transfer function is used. Algebraic differential operations are: And perform a Laplace transform on each term of equation (12), The Laplace transform is At the same time, taking into account and The term can be compensated for by adjusting the adjustable feedforward gain, ultimately yielding the frequency domain q-axis feedforward current. Function expression:

[0254] (14)

[0255] in, for Laplace transform of the time-speed command for Laplace transform of the q-axis feedforward current at time t.

[0256] To transform it into a difference equation that can be executed in a digital controller, intermediate variables are first introduced. Let it be: Equation (14) can then be written as:

[0257] (15)

[0258] For intermediate variables Cross-multiplication and rearranging the terms of the definition yields:

[0259] (16)

[0260] Convert it back to the time domain correspond This yields the first-order ordinary differential equation satisfied by the intermediate variables:

[0261] (17)

[0262] In the Each control cycle is discretized using the forward Euler method, and the sampling time is... Continuous variables and The sampled values ​​at that moment were respectively and In the first Each control cycle is represented by the forward difference instead of the differential. After substituting and rearranging, the solution is:

[0263] (18)

[0264] Through the discretization process described above, the continuous intermediate variables in the original continuous domain differential equation are... Transformed into discrete intermediate variables And thus, a difference equation that can be recursively computed in a digital signal processor was obtained. When the system powers on, and Initialization. The digital signal processor only needs to store the data from the previous cycle in each control cycle. Compared to the previous cycle Read the current Then, update recursively according to formula (18). .

[0265] In the frequency domain, the expression for the q-axis feedforward current is equation (15). In the time domain, Derivative of the intermediate variable , correspond Time and speed command , Corresponding constant Converting the above frequency domain expression back to the time domain yields: In the Each control cycle is represented by the forward difference instead of the differential. Continuous variables The sampled value at that moment After sorting, the first... q-axis feedforward current per control cycle expression:

[0266] (19)

[0267] In the The first control cycle, the... Speed ​​deviation per control cycle Expressed as:

[0268] (20)

[0269] in, For the first Speed ​​command per control cycle For the first Control cycle Calibrate the mechanical angular velocity.

[0270] The digital signal processor maintains a speed loop integral accumulation variable in memory, denoted as the th . The value updated after each control cycle Its recursive calculation formula is:

[0271] (twenty one)

[0272] in, For the first The value updated after each control cycle is initialized to [value] when the system powers on. .

[0273] Finally, the speed loop PI controller outputs the q-axis current reference component based on the speed deviation. :

[0274] (twenty two)

[0275] in, This is the proportional gain of the speed loop. This is the integral gain of the speed loop.

[0276] The first q-axis feedforward current per control cycle The q-axis current reference component generated by the speed loop PI controller based on the speed deviation Superimposed, forming the final first The q-axis given current in each control cycle :

[0277] (twenty three)

[0278] The compensation effect of the q-axis feedforward current is mainly reflected in the frequency range of speed command changes. In steady-state and gradually changing processes, the speed command changes slowly, and the q-axis feedforward current provides torque compensation proportional to acceleration, reducing the dynamic burden on the speed loop PI controller. In the high-frequency range, due to the bandwidth-limited differentiator gain automatically rolling off, the q-axis feedforward current hardly responds to the high-frequency noise components in the command, preventing noise from being amplified and injected into the current loop through the feedforward channel. This frequency characteristic allows feedforward compensation to balance dynamic response speed and signal smoothness, providing sufficient compensation in the low-frequency range where fast response is required, and automatically attenuating in the high-frequency range where noise suppression is required.

[0279] Step 3: Magnetic flux linkage domain MRAS identification, construction of reference model and adjustable model, using the output of Step 1. The reference flux linkage is calculated online based on the d-axis and q-axis voltage commands output by the PI controller in the current loop of the MRAS identification unit, which calibrates the electric angular velocity and flux linkage domain. The adjustable flux linkage is calculated based on the inverse estimate of the inductance, the flux linkage-inductance ratio estimate, and the d-axis and q-axis current feedback values. The deviation between the reference flux linkage and the adjustable flux linkage is converted into a parameter correction value after decoupling and filtering.

[0280] The goal of this step is to identify the equivalent inductance of two physical parameters online. and equivalent permanent magnet flux .in, The equivalent inductance, measured in Henry (H), characterizes the ability of a winding to impede changes in current and determines how fast the current can change when a voltage is applied. The equivalent permanent magnet flux linkage, measured in Weber (Wb), characterizes the strength of the magnetic field of a permanent magnet and determines the torque generated per unit current.

[0281] In a permanent magnet synchronous motor, the equivalent inductance and equivalent permanent magnet flux Numerically, there are usually significant differences, with a difference of one or even several orders of magnitude. If directly compared in MRAS... and When two parameters are identified online, the corresponding error gradients and parameter update step sizes will be mismatched due to the severe asymmetry in the numerical scale of the two parameters. When adjusting the same adaptive gain, the parameter channel with the larger value will respond too aggressively and easily generate violent oscillations, while the parameter channel with the smaller value will respond slowly and remain almost motionless, causing the two parameters to fail to converge synchronously, making the system extremely difficult to tune.

[0282] To address this scale inconsistency issue, this invention mathematically reconstructs the parameters to be identified. Two new intermediate identification parameters are defined: the reciprocal of the inductance. Compared to magnetic flux inductance The formula is:

[0283] (twenty four)

[0284] in, It is the reciprocal of the inductance, and the unit is per henry (1 / H). This is the flux linkage inductance ratio, measured in amperes (A). Estimated using typical values: If ,but ;like ,but After conversion, and Typical values ​​fall within 10 3 and 10 2 The magnitude and scale difference are reduced from a hundred times to ten times, and the two parameter channels can use similar adaptive gains, reducing the difficulty of tuning.

[0285] During the online identification process, the system maintains the intermediate parameter, the reciprocal of the inductance. Compared to magnetic flux inductance The estimated value is denoted as the inverse estimate of the inductance. Estimated value of flux linkage inductance In the first The system uses the first control cycle. Inductance reciprocal estimate for each control cycle and the Estimated flux inductance ratio for each control cycle Perform flux linkage calculation, and then update the result based on error feedback to obtain the first... Inductance reciprocal estimate for each control cycle and the Estimated flux inductance ratio for each control cycle When actual physical parameters are needed, the estimated equivalent inductance can be obtained through inverse calculation. and equivalent permanent magnet flux estimates :

[0286] (25)

[0287] in, The equivalent inductance is estimated by... The reciprocal is used to obtain the result, with the unit being Henry (H). The equivalent permanent magnet flux linkage is estimated by... Divide by The result is given in Weber (Wb).

[0288] Before constructing the reference model, we must first explain the two input quantities of the reference model: the d-axis voltage command. and q-axis voltage command The generation process. These two voltage commands are generated by the current loop PI controller, in the first... The first control cycle, for the d-axis, the first... The d-axis given current in each control cycle In the non-MTPA operating condition used in this invention, the value is set to zero. d-axis current feedback value per control cycle The current is measured by a current sensor and obtained through coordinate transformation. The d-axis current loop PI controller calculates the first... The d-axis given current in the first control cycle and the first control cycle The difference in the d-axis feedback current during the first control cycle is used to obtain the first... d-axis current error per control cycle :

[0289] (26)

[0290] The d-axis current loop PI controller performs proportional-integral calculations on this error and outputs the d-axis voltage command. : , This represents the d-axis current error; in a digital controller, the integral term is implemented using discrete accumulation, denoted as... For the first The cumulative value of the d-axis current error over each control cycle is derived using the following recursive formula: initial value Then the first D-axis voltage command for each control cycle for:

[0291] (27)

[0292] in, This is the d-axis current loop proportional gain. This is the integral gain of the d-axis current loop.

[0293] For the q-axis, the first The q-axis given current in each control cycle It consists of two superimposed parts. The first part is the q-axis current reference component generated by the speed loop PI controller based on the speed deviation. The second part is the q-axis feedforward current per control cycle The sum of the two is equation (23) in step two: Calculation of the q-axis current loop PI controller The q-axis given current in each control cycle With the q-axis current feedback value per control cycle The difference is used to obtain the first... q-axis current error per control cycle :

[0294] (28)

[0295] After proportional-integral calculation, the q-axis voltage command is output. : In digital controllers, The error is the q-axis current; the integral term is implemented using discrete accumulation, denoted as... For the first The cumulative value of the q-axis current error over each control cycle is derived using the following formula: initial value Then the first Q-axis voltage command per control cycle for:

[0296] (29)

[0297] in, This is the proportional gain of the q-axis current loop. This is the integral gain of the q-axis current loop.

[0298] The above d-axis current loop proportional gain d-axis current loop integral gain q-axis current loop proportional gain q-axis current loop integral gain In step four, the equivalent inductance and equivalent permanent magnet flux identified by MRAS will be updated online (see equations (50) and (51)).

[0299] Thus, in each control cycle, the current loop PI controller outputs the d-axis voltage command. and q-axis voltage command These two voltage commands, on the one hand, drive the inverter and control the motor operation in the vector control closed loop after inverse Park transformation and SVPWM modulation; on the other hand, they are also fed as inputs into the reference model below to participate in the online calculation of stator flux linkage.

[0300] Then, a reference model and an adjustable model are constructed respectively.

[0301] The reference model is based on the stator voltage equations and does not depend on any parameters to be identified, making it a neutral identification benchmark. In the synchronously rotating dq coordinate system, the aforementioned d-axis voltage command is used. q-axis voltage command d-axis current feedback value q-axis current feedback value and the result obtained in step one Calibrate electrical angular velocity The stator flux linkage can be reconstructed online using the stator voltage equation:

[0302] (30)

[0303] in, This is the stator resistance, measured in ohms (Ω). The result obtained in step one Calibrate the electrical angular velocity, in radians per second (rad / s). d-axis reference flux linkage Here is the q-axis reference flux linkage, in Weber (Wb). The right-hand side of the equation... and The two terms are rotating back electromotive force terms, reflecting the cross-coupling caused by coordinate rotation. Additionally, The calibrated electrical angular velocity is used to characterize the rotational coupling relationship in the reference model, which can reduce the impact of encoder angular velocity noise on the reference flux reconstruction process; the d-axis voltage command and q-axis voltage command are respectively output by the current loop PI controller based on the error between the given d-axis and q-axis currents and the corresponding d-axis and q-axis current feedback values, and are correlated with... The calibrated electric angular velocity and d-axis and q-axis current feedback values ​​are used together to calculate the reference flux linkage of the reference model. The d-axis and q-axis voltage commands and current feedback values ​​characterize the impact of the current loop dynamics on the reference flux linkage reconstruction. Therefore, the source reconstruction unit and the command preset unit improve the reference flux linkage reconstruction conditions from the angular velocity signal source and the current command source, respectively, thereby reducing high-frequency noise and transient bias in the subsequent flux linkage error.

[0304] To implement this in a digital controller, in the first... In the first control cycle, the forward Euler method is used to discretize equation (30). Taking the d-axis as an example, in the first control cycle... One control cycle, in equation (30) respectively The differential is replaced by the forward difference: Substituting into the d-axis voltage equation, we get: Multiply both sides by Then Moving to the right, we obtain the discrete update formula for the d-axis reference flux:

[0305] (31)

[0306] Similarly, the discrete update formula for the q-axis reference flux can be obtained:

[0307] (32)

[0308] in, For the first Periodic d-axis reference flux and For the first Periodic q-axis reference flux; For the first Periodic d-axis reference flux and For the first The periodic q-axis reference flux is initially set to zero when the system is powered on.

[0309] Equations (31) and (32) mean that in the first... In each control cycle, the new value of the reference flux is equal to the old value retained from the previous cycle plus each term on the right-hand side of the voltage equation for the current cycle multiplied by the increment of the control cycle. By accumulating this increment in each cycle, the reference flux is gradually updated over time, which can better reflect the change process of the stator flux.

[0310] The parameters used in the adjustable model are the inverse estimates of the inductance. Estimated value of flux linkage inductance If the estimation is accurate, the calculated results will be consistent with the reference model; if the estimation is inaccurate, there will be deviations. In non-MTPA operating conditions ( Under these conditions, the d-axis current is zero, and its secondary effects on magnetic circuit saturation and d- and q-axis cross-coupling can be ignored. Therefore, this invention uses a single equivalent inductor. and equivalent permanent magnet flux This describes the relationship between stator flux linkage and current. The original flux linkage equation is:

[0311] (33)

[0312] Substitute parameters and Substituting the values, we obtain the d-axis adjustable flux linkage. : By combining the two terms on the right with a common denominator of b, we obtain the adjustable flux linkage along the d-axis. :

[0313] (34)

[0314] Similarly, the q-axis adjustable flux linkage is obtained. :

[0315] (35)

[0316] In the Each control cycle will display the actual parameters. , Replace with the first Inductance reciprocal estimate for each control cycle and the Estimated flux inductance ratio for each control cycle , and combined with the first d-axis current feedback value per control cycle and the q-axis current feedback value per control cycle The adjustable flux linkages along the d and q axes are obtained as follows:

[0317] (36)

[0318] in, For the first Adjustable d-axis flux for each cycle For the first The q-axis adjustable flux linkage for each cycle, in Weber (Wb). For the first d-axis current feedback value per control cycle No. q-axis current feedback value for each control cycle; For the first The inverse estimate of the inductance per control cycle. For the first Estimated flux linkage inductance ratio for each control cycle.

[0319] Therefore, within a control cycle, the system simultaneously possesses the reference flux provided by the reference model and the adjustable flux provided by the adjustable model. The difference between the two is the basis for updating the driving parameters, which will be elaborated in step four.

[0320] Step 4: Gain scheduling Figure 6 The cumulative correction value is generated based on the shaping drive signal. The cumulative correction value is superimposed on the initial intermediate parameter estimate to obtain the updated inverse inductance estimate and flux-inductance ratio estimate. The equivalent inductance and equivalent permanent magnet flux are calculated in reverse. The current loop PI controller parameters are updated in real time through the gain scheduling strategy. The updated inverse inductance estimate and flux-inductance ratio estimate are fed back to the adjustable model for use in calculating the adjustable flux in the next control cycle.

[0321] In the In each control cycle, the components of the reference flux and the adjustable flux on the d and q axes are subtracted to obtain the flux error:

[0322] (37)

[0323] in, For the first The d-axis flux error per control cycle For the first The q-axis flux linkage error per control cycle is expressed in Weber (Wb). These two errors collectively reflect... and The degree of deviation from the truth, when and When the value is zero, the identification is complete.

[0324] but and It simultaneously includes parameter deviation information and current operating condition information, requiring the construction of corresponding parameters based on parameter sensitivity relationships. and The instantaneous drive signal is constructed based on the design criteria of Popov's superstability theory. To ensure asymptotic stability in the parameter adaptive update process, the instantaneous drive signal should be constructed by combining flux linkage error and current feedback signals, and together with the subsequent linear forward channel and PI adaptive law, satisfy the strict positive reality design requirements. Based on the above principles, the instantaneous drive signal is constructed as follows:

[0325] (38)

[0326] in, For the first Driven by one control cycle The instantaneous drive signal integrates the flux linkage error of the d and q axes as well as current information; its amplitude and sign indicate... The direction and magnitude of the deviation; For the first Driven by one control cycle The instantaneous drive signal is directly used with respect to the d-axis flux linkage error, because The change in flux mainly affects the amplitude of the d-axis flux linkage. Under vector control conditions where the d-axis current is zero, Typically close to zero, the two instantaneous drive signal channels have good decoupling characteristics.

[0327] and The parameter estimate is obtained by multiplying and adding the flux linkage error and the instantaneous current value, inevitably carrying current sampling noise and encoder residual ripple. If these instantaneous drive signals are directly fed into the adaptive law, the parameter estimate will fluctuate or even oscillate at high frequency. Therefore, this invention inserts a first-order low-pass forward channel between the error signal and the adaptive update law to perform dynamic shaping.

[0328] In the continuous domain, the differential equation of a first-order low-pass filter is:

[0329] (39)

[0330] (40)

[0331] in, for The forward channel gain of a channel controls the filtering bandwidth of that channel and determines... For instantaneous error signals The balance between tracking speed and high-frequency suppression capability; for The forward channel gain of a channel controls the filtering bandwidth of that channel and determines... For instantaneous error signals The balance between tracking speed and high-frequency suppression capability.

[0332] To implement this in a digital controller, forward Euler method is used to discretize equations (39) and (40). For example, a channel:

[0333] The differential can be replaced by the forward difference: Substitution formula (39):

[0334] (41)

[0335] Final result:

[0336] (42)

[0337] Similarly, we can obtain The recursive formula for the channel:

[0338] (43)

[0339] in, For the first After one cycle of plastic surgery Shaping drive signal, For the first After one cycle of plastic surgery Shaping drive signal. Equations (42) and (43) mean that in each control cycle, the new output value is equal to the decay of the retained value from the previous cycle plus the weighted sum of the new input value for the current cycle. Gain , The larger the gain, the faster the attenuation, the larger the channel bandwidth, and the more responsive the response, but the weaker the noise suppression; the smaller the gain, the better. By selecting the appropriate gain... , , making , It can quickly track error trends without being swayed by individual sampling spikes.

[0340] Obtained shaping drive signal and Then, they need to be converted to pairs. and The cumulative correction amount. This invention employs a proportional-integral adaptive update law to obtain the cumulative correction amount over the continuous domain. and expression:

[0341] (44)

[0342] (45)

[0343] in, for Channel proportional gain, for Channel proportional gain, for Channel integral gain, for Channel integral gain.

[0344] In the For each control cycle, the proportional terms in equations (44) and (45) are directly taken as their current values, and the integral terms are replaced by discrete summation: , , obtain the discrete domain up to the th Cumulative correction amount per control cycle and expression:

[0345] (46)

[0346] (47)

[0347] in As of the date The cumulative correction to the inverse estimate of the inductance over each control cycle. As of the date The cumulative correction to the estimated flux-inductance ratio over each control cycle. For summation index variables.

[0348] The role of the PI structure here is that in the early stage of identification, when the parameter deviation is large and the driving signal amplitude is large, the proportional term provides a large step correction to quickly reduce the deviation; as the parameters approach the true value, the driving signal gradually decreases, the contribution of the proportional term shrinks, and at this time the integral term has accumulated enough historical information to continuously output a stable and gentle correction force, which gradually eliminates the residual small steady-state error, which is conducive to reducing the steady-state parameter deviation.

[0349] Finally, the number was obtained The cumulative correction for each control cycle is then added to the initial estimate to obtain the parameter estimate for the next control cycle.

[0350] (48)

[0351] in, For the first Inductance reciprocal estimate for each control cycle For the first The estimated flux linkage-to-inductance ratio for the first control cycle will be... The new current value of each control cycle is substituted into the adjustable model equation (36) to perform a new round of flux linkage calculation, forming a closed-loop iteration. Its update mechanism can be summarized as follows: In each control cycle, the system calculates the adjustable flux linkage based on the current parameter estimate and compares it with the reference flux linkage to obtain the flux linkage error; after dynamic shaping and proportional-integral adaptive update law, the flux linkage error forms the cumulative correction up to the current control cycle. The system uses the initial estimate as a benchmark and adds this cumulative correction to obtain the parameter estimate for the next control cycle, and substitutes it into the adjustable model equation (36) as the new current value to perform a new round of flux linkage calculation, forming a closed-loop iteration. As the iteration progresses, the adjustable flux linkage and the reference flux linkage tend to be consistent, and the parameter estimate gradually approaches the actual parameters. In the first control cycle... After the intermediate parameters are updated in each control cycle, the inverse calculation relationship defined in formula (25) in step three is used to... , Convert to , The calculation formulas are as follows: , .

[0352] The identified parameters are used to update the control loop parameters in real time. This invention employs a gain scheduling strategy to map the identification results of the MRAS to the parameters of the current loop PI controller, achieving online tuning. First, based on the identified equivalent permanent magnet flux linkage... The target bandwidth of the current loop is calculated using the magnitude of the current.

[0353] (49)

[0354] in, The target bandwidth of the current loop is expressed in radians per second (rad / s). The base bandwidth is the minimum bandwidth when the flux linkage is zero. To avoid excessively large target bandwidth causing current loop oscillations, in practical implementations, it can be adjusted... Set the upper limit value; This represents the flux linkage modulation coefficient. When the identified flux linkage is strong, the target bandwidth is increased accordingly to fully utilize the motor's torque output capability.

[0355] Then, in the... Each control cycle is based on the target bandwidth and the identified equivalent inductance estimate. Update the current loop PI controller parameters according to the standard first-order current model design principles. For the d-axis current loop:

[0356] (50)

[0357] For the q-axis current loop:

[0358] (51)

[0359] in, This is the d-axis current loop proportional gain. This is the d-axis current loop integral gain; This is the proportional gain of the q-axis current loop. This is the integral gain of the q-axis current loop; This is the stator resistance.

[0360] Figure 8 The graph shows a comparison of the speed response and current reference with and without source reconstruction. The main curve in the graph compares the overall trends of the electrical angular velocity calibrated by the source reconstruction unit, the coarse estimate of the electrical angular velocity, the current reference signal optimized by the source reconstruction unit, and the unoptimized current reference signal. As can be seen from the main graph, during the acceleration phase, the overall trends of the two speed signals are basically the same, indicating that the source reconstruction unit does not change the basic dynamic process of the motor speed response. After entering steady state, the unoptimized coarse estimate of the electrical angular velocity exhibits significant high-frequency fluctuations, while the calibrated electrical angular velocity processed by the source reconstruction unit is smoother.

[0361] The enlarged area in the lower left corner of the figure shows the details of steady-state speed fluctuations. The unoptimized speed signal has a root mean square (RMS) fluctuation of 0.1834 rad / s and a peak-to-peak (peak-to-peak) fluctuation of 0.56 rad / s. After optimization by the source reconstruction unit, the RMS fluctuation of the speed is reduced to 0.0228 rad / s, and the peak-to-peak fluctuation is reduced to 0.091 rad / s. The enlarged area in the lower right corner of the figure shows the details of current reference signal fluctuations. The unoptimized current reference signal has a RMS fluctuation of 0.2884 A and a peak-to-peak (peak-to-peak) fluctuation of 1.36 A. After optimization by the source reconstruction unit, the RMS fluctuation of the current reference signal is reduced to 0.2223 A, and the peak-to-peak fluctuation is reduced to 0.96 A.

[0362] The above results show that the source reconstruction unit can effectively reduce high-frequency noise in the coarse estimation signal of electric angular velocity and further reduce the current reference ripple caused by speed feedback fluctuations, so that the subsequent flux linkage domain MRAS identification unit can obtain a smoother and more stable input signal, which is conducive to improving the stability and noise resistance of the parameter identification process.

[0363] Experimental verification

[0364] Before the experiment, the back electromotive force characteristics of the motor were measured. The motor was externally driven to 1000 rpm, and the terminal voltage signal was acquired using an oscilloscope. The results are as follows. Figure 9 As shown, the back electromotive force waveform has an approximately sinusoidal distribution, indicating that the electromagnetic structure of the motor has good symmetry. Statistical analysis shows that its effective value is approximately 9.27 V, based on which the safe operating speed range of the system is determined.

[0365] To verify the effectiveness of the method proposed in this invention, the performance of two control structures was compared under step speed command and sinusoidal speed command: the first is the basic MRAS identification structure without the addition of the source reconstruction unit and the command preset unit; the second is the MRAS identification structure with the addition of the source reconstruction unit and the command preset unit.

[0366] To verify the impact of source reconstruction and command presetting on the stability of parameter identification, the online estimation results of equivalent parameters b and c under three different MRAS identification structures were compared at a speed of 1000 rpm. The results are as follows: Figure 10As shown in the figure, for parameter b, the basic MRAS identification structure exhibits significant transient overshoot and subsequent oscillations during startup and convergence. The estimated parameter value fluctuates around the steady-state value of 338.932, with large peak-to-peak fluctuations. After adding the source reconstruction unit, the initial overshoot amplitude of parameter b is reduced, but significant low-frequency fluctuations still exist during subsequent operation. This indicates that source reconstruction alone can improve the signal source quality, but it cannot fully suppress the dynamic oscillations during identification. With the addition of both the source reconstruction unit and the instruction preset unit, parameter b can reach the vicinity of the steady-state value more quickly. The convergence time advance indicated in the figure is approximately ΔTb = 0.460s, and the steady-state fluctuation amplitude is significantly reduced. This shows that the instruction preset unit can further improve the impact of dynamic current tracking error on parameter identification. For parameter c, the estimation result under the basic MRAS identification structure exhibits significant periodic fluctuations around the steady-state value of -28.98. Adding the source reconstruction unit improves the fluctuation of parameter c, but some oscillations still exist. Simultaneously adding both the source reconstruction unit and the instruction preset unit results in a smoother estimation curve for parameter c, which remains relatively stable near the steady-state value. Combining the estimation results of parameters b and c, it can be seen that the source reconstruction unit mainly suppresses the influence of encoder feedback noise and angle inconsistency on the identification input, while the instruction preset unit further reduces dynamic current errors during acceleration and deceleration. Their synergistic effect improves the convergence speed, steady-state smoothness, and overall reliability of MRAS parameter identification. This result shows that adding the source reconstruction unit alone can reduce transient peak values ​​caused by angular velocity noise, but low-frequency fluctuations still exist during the dynamic process. Further adding the instruction preset unit improves both the response speed and steady-state fluctuations of parameter estimation, indicating that the two units have a synergistic suppression effect at the source of MRAS error signals, rather than a simple superposition of the effects of a single module.

[0367] Figure 11 and Figure 12 The d-axis and q-axis current response waveforms and speed response waveforms of the motor under a step speed command are given when the source reconstruction unit and the command preset unit are not added. Figure 11 In the figure, the horizontal axis represents time t, in seconds (s), and the vertical axis represents current, in amperes (A); This represents the given current along the d-axis. This represents the d-axis current feedback value. This represents the given current along the q-axis. This represents the q-axis current feedback value. This represents the positive peak value of the d-axis current feedback. This represents the negative peak value of the d-axis current feedback. Figure 11 It can be seen that, without the addition of the source reconstruction unit and the instruction preset unit, during the step speed change process... A significant impact occurred, with its positive peak value. Reaching 30.2329A, negative peak value It is -15.5156 A; at the same time, In tracking The process exhibits certain fluctuations, indicating that the current tracking error is quite significant during step speed change.

[0368] Figure 12 In the figure, the horizontal axis represents time t, in seconds (s), and the vertical axis represents rotational speed, in Hertz (Hz). Indicates the speed reference signal. This represents the speed feedback signal. Without the addition of the signal source reconstruction unit and the instruction preset unit, the peak-to-peak speed fluctuations of the motor in the steady-state stages of 80 Hz and 100 Hz are 4.2297 Hz and 5.067 Hz, respectively, indicating that the steady-state speed fluctuations are relatively large under this control structure.

[0369] Figure 13 and Figure 14 The d-axis current response waveforms and speed response waveforms of the motor under a step speed command are given after the addition of the source reconstruction unit and the command preset unit. Figure 13 middle, This represents the given current along the d-axis. This represents the d-axis current feedback value. This represents the given current along the q-axis. This represents the q-axis current feedback value. This represents the positive peak value of the d-axis current feedback. This represents the negative peak value of the d-axis current feedback. Figure 13 It can be seen that after adding the source reconstruction unit and the instruction preset unit, the step condition... The impact amplitude decreased. Decreased to 24.503 A. It is -15.876 A; at the same time, and The tracking relationship is more stable, and the fluctuation amplitude between the two is reduced, indicating that the instruction preset unit can improve the dynamic tracking performance of the q-axis current during the step speed change process.

[0370] Figure 14 In the diagram, after adding the source reconstruction unit and the instruction preset unit, the... Indicates the speed reference signal. This indicates that the peak-to-peak speed fluctuations of the motor in the steady-state phases of 80 Hz and 100 Hz have decreased to 3.5298 Hz and 4.014 Hz, respectively. Combined with... Figures 11 to 14It can be seen that under the step speed command, the source reconstruction unit and the command preset unit work together to reduce the d-axis current impact, improve the q-axis current tracking relationship, and reduce steady-state speed fluctuations.

[0371] Figure 15 and Figure 16 The d-axis current response waveforms and speed response waveforms of the motor under sinusoidal speed commands are given when the source reconstruction unit and command preset unit are not added. Figure 15 In the figure, the horizontal axis represents time t, in seconds (s), and the vertical axis represents current, in amperes (A); This represents the given current along the d-axis. This represents the d-axis current feedback value. This represents the given current along the q-axis. This represents the q-axis current feedback value. (From...) Figure 15 It can be seen that, without the addition of the source reconstruction unit and the instruction preset unit, the sinusoidal speed change process... follow There are changes, but there is a certain deviation between the two; at the same time, Compared to The presence of fluctuations indicates that there is still a certain coupling effect between the d-axis and q-axis currents under sinusoidal operating conditions.

[0372] Figure 16 In the figure, the horizontal axis represents time t, in seconds (s), and the vertical axis represents rotational speed, in Hertz (Hz); Indicates the speed reference signal. This indicates the speed feedback signal. This indicates the tracking time difference between the speed feedback signal and the speed reference signal. This indicates the peak-to-peak value of the rotational speed fluctuation. (From...) Figure 16 It can be seen that without the addition of the source reconstruction unit and the instruction preset unit, Compared to There is a significant phase lag and a time difference in speed tracking. The peak-to-peak value of the rotational speed fluctuation is approximately 0.38 s. It is approximately 25.3334 Hz.

[0373] Figure 17 and Figure 18 The d-axis current response waveforms and speed response waveforms of the motor under sinusoidal speed commands are given after the addition of the source reconstruction unit and the command preset unit. Figure 17 In the middle, in the picture This represents the given current along the d-axis. This represents the d-axis current feedback value. This represents the given current along the q-axis. This represents the q-axis current feedback value. (From...) Figure 17It can be seen that after adding the source reconstruction unit and the instruction preset unit, right The tracking is more stable. Compared to The reduced fluctuation indicates that the d-axis and q-axis current coupling is suppressed under sinusoidal conditions, and the current tracking performance is improved.

[0374] Figure 18 In the middle, in the picture Indicates the speed reference signal. This indicates the speed feedback signal. This indicates the tracking time difference between the speed feedback signal and the speed reference signal. This indicates the peak-to-peak value of the rotational speed fluctuation. (From...) Figure 18 It can be seen that after adding the source reconstruction unit and the instruction preset unit, and The phase difference between them decreases, and the speed tracking time difference decreases. Shortened to approximately 0.27 s, peak-to-peak speed fluctuation Reduced to approximately 24.5938 Hz. Combined with... Figures 15 to 18 It can be seen that under sinusoidal speed command, the synergistic effect of the source reconstruction unit and the command preset unit can improve speed tracking lag, reduce speed fluctuation, and improve the stability of d-axis and q-axis current tracking.

[0375] The foregoing has shown and described the basic principles and main features of the present invention. Those skilled in the art should understand that the above embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention; various changes and modifications can be made by those skilled in the art without departing from the spirit and essence of the present invention, and all such changes and modifications should fall within the scope of protection of the present invention.

Claims

1. A method for identifying permanent magnet synchronous motors that combines source reconstruction and command preset coordination, applied to vector control conditions where the d-axis given current is zero, characterized in that... Includes the following steps: Step 1: Reconstruct the source signal of the raw electrical angle output by the encoder to obtain... Calibrate electric angular velocity and Internal electrical angle; Step 2: Generate q-axis feedforward current based on speed command and mechanical dynamics relationship, and superimpose the q-axis feedforward current with the speed loop PI output to form q-axis given current; Step 3: Based on the... A reference model for the flux linkage domain MRAS is constructed using calibrated electric angular velocity, d-axis voltage command, q-axis voltage command, d-axis current feedback value, and q-axis current feedback value, and based on the inverse inductance estimate... Estimated value of flux linkage inductance ratio An adjustable model is constructed using d-axis and q-axis current feedback values. , , This is an estimate of the equivalent inductance. The equivalent permanent magnet flux linkage is estimated; Step 4: Update the estimated values ​​of b and c based on the flux linkage error between the reference model and the adjustable model, back-calculate the estimated values ​​of the equivalent inductance and the equivalent permanent magnet flux linkage, and update the current loop PI controller parameters based on the estimated values ​​of the equivalent inductance and the equivalent permanent magnet flux linkage; wherein, the The calibrated electrical angular velocity is used to calculate the rotating back electromotive force term in the reference model. The d-axis voltage command and q-axis voltage command are respectively output by the current loop PI controller based on the error between the given current of the d and q axes and the corresponding d and q axis current feedback values, and are used for the reference flux calculation of the reference model, so that the encoder angular velocity noise and dynamic current tracking error during speed change are synchronously reduced before flux error is formed.

2. The permanent magnet synchronous motor identification method according to claim 1, characterized in that, Step one includes: [The text abruptly ends here, likely due to an incomplete sentence or a formatting error.] The original electrical angle of the first control cycle and the first control cycle The original electrical angle of the first control cycle is differentially calculated and processed by a wrapper function to obtain a coarse estimate of the electrical angular velocity; the first... The original electrical angle of the first control cycle and the first control cycle Each control cycle The phase error is obtained by subtracting the internal electrical angles and processing them using a wrapping function; a speed correction term is generated based on the phase error using a proportional-integral loop filter; the coarse estimate of the electrical angular velocity is added to the speed correction term to obtain... Calibrate the electric angular velocity, and according to the above... The next control cycle is obtained by recursively calculating the calibrated electrical angular velocity. Internal electrical angle.

3. The permanent magnet synchronous motor identification method according to claim 1, characterized in that, In step two, the q-axis feedforward current is determined based on the moment of inertia, the change in speed command, the viscous damping term, and the load torque term. When the load torque is not obtained, the load torque is set to 0, and the remaining load disturbance is compensated by the speed loop PI feedback channel and the adjustable feedforward gain. When the estimated load torque value is obtained, the estimated load torque value is substituted into the calculation of the q-axis feedforward current.

4. The permanent magnet synchronous motor identification method according to claim 1, characterized in that, In step three, the adjustable model is constructed based on the inverse estimate of the inductance, the estimate of the flux linkage-to-inductance ratio, and the d-axis and q-axis current feedback values, and satisfies: , ,in For the first One control cycle for the d-axis adjustable flux linkage For the first One control cycle for the q-axis adjustable flux linkage For the first q-axis current feedback value per control cycle For the first d-axis current feedback value per control cycle For the first Inductance reciprocal estimate for each control cycle For the first Estimated flux linkage inductance ratio for each control cycle.

5. The permanent magnet synchronous motor identification method according to claim 1, characterized in that, In step four, corresponding structures are constructed based on the d-axis and q-axis flux linkage errors. and The driving signal, wherein the d-axis and q-axis flux linkage errors are obtained by subtracting the reference flux linkage output by the reference model from the adjustable flux linkage output by the adjustable model; wherein, the reference model adopts... The reference flux linkage is calculated using the d-axis and q-axis voltage commands output by the current loop PI controller within the calibrated electric angular velocity and flux linkage domain MRAS identification unit, along with the d-axis and q-axis current feedback values. The d-axis voltage command is output by the current loop PI controller based on the error between the given d-axis current and the d-axis current feedback value, and the q-axis voltage command is output by the current loop PI controller based on the error between the given q-axis current and the q-axis current feedback value. The drive signal is then dynamically shaped via a first-order low-pass forward channel to obtain a shaped drive signal. and And generate up to the first according to the shaping drive signal. Cumulative correction amount per control cycle and The cumulative corrections are then added to the initial estimate. and To obtain the next control cycle and .

6. The permanent magnet synchronous motor identification method according to claim 5, characterized in that, The cumulative correction amount satisfies: ,in As of the date The cumulative correction to the inverse estimate of the inductance over each control cycle. As of the date The cumulative correction to the estimated flux-inductance ratio over each control cycle. For summation index variables, To control the cycle, for Channel proportional gain, for Channel proportional gain, for Channel integral gain, for Channel integral gain, For the first After one cycle of plastic surgery Shaping drive signal, For the first After one cycle of plastic surgery Shaping drive signal.

7. A permanent magnet synchronous motor identification device that coordinates source reconstruction and command preset, characterized in that, It includes a source reconstruction unit, an instruction preset unit, a flux linkage domain MRAS identification unit, and a gain scheduling unit; the source reconstruction unit is used to reconstruct the source of the original electrical angle output by the encoder, and output... Calibrate electric angular velocity and Internal electrical angle; the stated The calibrated electric angular velocity was obtained after pole pair conversion. The mechanical angular velocity is calibrated and used as the speed feedback input to the command preset unit; the command preset unit provides the d-axis and q-axis setpoint currents to the current loop PI controller in the flux linkage domain MRAS identification unit, wherein the q-axis setpoint current is formed by superimposing the q-axis feedforward current and the speed loop PI output, and the d-axis setpoint current is set to zero under the vector control condition where the d-axis setpoint current is zero; the flux linkage domain MRAS identification unit is used to... A reference model is constructed using the calibrated electric angular velocity, the d-axis and q-axis voltage commands output by the internal current loop PI controller, and the d- and q-axis current feedback values. An adjustable model is then constructed based on the inverse inductance estimate, the flux linkage-to-inductance ratio estimate, and the d- and q-axis current feedback values. This aims to simultaneously reduce the impact of angular velocity noise and dynamic current tracking errors on parameter identification before flux linkage errors occur. The d-axis and q-axis voltage commands are output by the current loop PI controller based on the errors between the given d- and q-axis currents and their corresponding d- and q-axis current feedback values. The gain scheduling unit updates the estimates of b and c based on the flux linkage error between the reference model and the adjustable model, inversely calculates the equivalent inductance estimate and the equivalent permanent magnet flux linkage estimate, and updates the current loop PI controller parameters.

Images