A dual-rate model-free control system and method driven by a high-frequency motor

Abstract

This invention discloses a dual-rate model-free control method for high-frequency motor drives. Its core principle lies in separating computationally intensive tasks from high-rate control execution. Specifically, a complex online model identification and delay compensation are performed through a slow-rate interrupt to obtain an accurate, data-driven model that requires no physical parameters. Simultaneously, an interpolation time-shift prediction method is used to generate a control sequence for the high-rate control cycle within the slow-rate interrupt. Subsequently, the high-rate interrupt is only responsible for executing this pre-calculated control sequence sequentially. Since correction based on the latest identified model is implemented at each high-rate control point, and the computational burden of the high-rate interrupt is extremely low, this method fundamentally solves the real-time constraints and model parameter dependencies in high-frequency control, thus achieving highly robust quasi-cycle-by-cycle control of the high-frequency motor drive system.

Description

Technical Field

[0001] This invention belongs to the field of high-frequency motor drive control technology, and more specifically, relates to a dual-rate model-free control system and method for high-frequency motor drive. Background Technology

[0002] The development of wide-bandgap semiconductor devices has enabled high-frequency inverters to offer lower switching losses and faster switching speeds, leading to their widespread application in aerospace, high-speed drones, and electric vehicles. These advantages have made it possible to improve the power density of motor drive systems, reduce current ripple, and increase control bandwidth. However, the extremely high switching frequency compresses the interrupt cycle of the digital controller, making it difficult for the digital controller to complete complex, high-performance cycle-by-cycle control algorithms in a very short time. To address this challenge, a model-based dual-rate predictive current control scheme has emerged. This scheme utilizes an improved model at a slow sampling rate to predict missing fast output data, allowing the control frequency to exceed the sampling frequency, thus increasing the switching frequency while alleviating the computational burden on the processor.

[0003] However, although the aforementioned model-based dual-rate schemes achieve high-frequency control, their control performance is highly dependent on an accurate prior physical model of the system. Under high-frequency conditions, parameter mismatch and unmodeled nonlinear dynamics lead to model inaccuracies, and the prediction error accumulates during the prediction process, thus degrading control performance. To eliminate dependence on prior physical models, model-free control strategies have been proposed, employing data-driven models to identify and describe system dynamics in real time. However, existing model-free control schemes are mainly designed for traditional single-rate frameworks and fail to address the problem of missing output data caused by the mismatch between the sampling rate and control rate in dual-rate systems. This hinders the application of model-free control in high-frequency dual-rate systems. Therefore, a novel control method capable of achieving model-free dual-rate control is urgently needed. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and propose a dual-rate model-free control system and method for high-frequency motor drive. This solves the problem of missing output data faced by model-free control in the dual-rate framework. By decoupling the slow-rate data-driven modeling and the fast-rate control execution, a high switching frequency is achieved while ensuring the system has excellent robust performance.

[0005] To achieve the above-mentioned objectives, the present invention provides a dual-rate model-free control system for high-frequency motor drive, characterized in that it comprises: a three-phase permanent magnet synchronous motor (PMSM) 1, a silicon carbide-based high-frequency driver 2, a DC power supply 3, a position encoding module 4, a current sampling module 5, a speed calculation module 6, a difference operation unit 7, and a speed loop PI controller 8. / Coordinate transformation module 9, space vector pulse width modulation module 10, and dual-rate model-free controller 11;

[0006] The three-phase permanent magnet synchronous motor 1 is the system control object, and its energy input comes from the three-phase AC power of the high-frequency driver 2;

[0007] The high-frequency motor driver 2 is a three-phase voltage source inverter VSI, with its input side connected to a DC power supply 3 and its output side connected to a three-phase permanent magnet synchronous motor 1.

[0008] The DC power supply 3 is the DC voltage source of the system, providing a stable DC voltage for the high-frequency driver 2;

[0009] The position encoding module 4 is a position detection device, which is installed on the shaft of the three-phase permanent magnet synchronous motor 1 and is used to detect electrical angles. And pass it to the speed calculation module 6;

[0010] The current sampling module 5 is a current detection device, which is installed between the high-frequency driver 2 and the permanent magnet synchronous motor 1 to collect three-phase AC current. And pass it on Coordinate transformation module 9;

[0011] The speed calculation module 6 is connected to the position encoding module 4, and calculates the speed based on the electrical angle. The angular velocity of the motor is calculated in real time. ;

[0012] The difference operation unit 7 is an error calculation unit, and its input comes from the reference angular velocity. Angular velocity of velocity calculation module 6 The speed error is obtained by calculating the difference between the two.

[0013] The input to the speed loop PI controller 8 is the speed error of the difference calculation unit 7, which is obtained through the proportional-integral controller. Shaft reference current ;

[0014] The / Coordinate transformation module 9 is a spatial coordinate transformation matrix, which transforms the three-phase currents respectively. Convert to Current in stationary coordinate system general shaft reference current converted to Reference current in stationary coordinate system ;

[0015] The dual-rate model-free controller 11 receives from / The current signal from coordinate transformation module 9 is processed by an internal algorithm and then outputs a high-frequency control sequence to space vector pulse width modulation module 10.

[0016] The input terminal of the space vector pulse width modulation module 10 is connected to the dual-rate model-free controller 11, and is used to process the obtained high-frequency control sequence. The signals are modulated sequentially to generate the high-frequency switching signals required to drive the high-frequency driver 2. .

[0017] In addition, the present invention also provides a dual-rate model-free control method for high-frequency motor drive, characterized by comprising the following steps:

[0018] (1) Slow-rate task: data sampling, latency compensation and model identification. This step is performed during the slow-rate sampling period. Internal execution:

[0019] (1.1) Data sampling of the motor system: The three-phase current of the motor is obtained through the current sampling module. and through The coordinate transformation module converts it to Actual current in stationary coordinate system , For slow sampling time, It is a positive integer;

[0020] (1.2) Time delay compensation of digital controller: A hyperlocal model is used to extend the state observer to the current sampled current. Perform filtering and predict the next slow-rate current. This is used as the starting point for deducing the control input sequence to reduce the impact of computational delay.

[0021] (1.3) Data-driven model identification: The autoregressive moving average (ARMAX) model is used to describe the dynamics of the motor system. The recursive least squares (RLS) algorithm is used to identify the model based on the slow-speed input voltage. and output current Online identification and updating of the coefficient vector of ARMAX models ;

[0022] (2) High-rate task: Constructing the control law, reconstructing missing data, and generating the control sequence. This step is calculated during the slow-rate interrupt and is part of the high-rate cycle. Generate a set of control sequences:

[0023] (2.1) Constructing the fast rate control law: Based on the ARMAX model coefficients identified in step (1.3) A time-shift prediction strategy is adopted to construct a control law equation suitable for fast-rate cycles;

[0024] (2.2) Reconstruction of missing output data: due to slow sampling points and Between, high rate of output The missing information is addressed in this step using interpolation, which utilizes known information. and prediction To estimate the missing intermediate data points;

[0025] (2.3) Solving for the optimal control sequence: Substitute the missing data reconstructed in step (2.2) into the fast rate control law in step (2.1), and combine it with the current reference command to recursively solve for the optimal control sequence containing... Optimal control sequence for each control input ;

[0026] (3) Drive execution: Quasi-cycle-by-cycle control of the control sequence is achieved by adopting a dual-rate interrupt nesting mechanism:

[0027] (3.1) Low-rate interruption (sampling frequency) ): The computationally intensive tasks of sampling, delay compensation, and RLS model identification in step (1) and the control sequence generation task in step (2) are executed sequentially.

[0028] (3.2) High-speed interrupt (switching frequency) This interrupt is only responsible for performing simple output tasks in each high-rate control cycle. Initially, a control sequence is sequentially selected from the pre-calculated optimal control sequence of the low-rate interrupt. The signal is input to the Space Vector Pulse Width Modulation (SVPWM) module and generates a high-frequency switching signal.

[0029] The objective of this invention is achieved as follows:

[0030] The dual-rate model-free control method for high-frequency motor drives proposed in this invention is based on the core principle of separating computationally intensive tasks from high-rate control execution. Specifically, a complex online model identification and delay compensation are performed through a slow-rate interrupt to obtain an accurate, data-driven model that does not require physical parameters. Simultaneously, an interpolation time-shift prediction method is used to generate the control sequence of the high-rate control cycle in one go within the slow-rate interrupt. Subsequently, the high-rate interrupt is only responsible for executing the pre-calculated control sequence sequentially. Since the correction based on the latest identified model is implemented at each high-rate control point, and the computational burden of the high-rate interrupt is extremely low, this method fundamentally solves the problems of computational real-time constraints and model parameter dependence in high-frequency control, thus achieving highly robust quasi-cycle-by-cycle control of the high-frequency motor drive system.

[0031] Meanwhile, the dual-rate model-free control method based on high-frequency motor drive of the present invention also has the following beneficial effects:

[0032] (1) By decoupling the sampling and control frequency, the present invention places the computationally intensive model identification and control law solving tasks in a slow-rate interrupt, while the high-rate interrupt is only responsible for the cycle-by-cycle execution of instructions. This method effectively alleviates the computational pressure on the digital processor under high switching frequency, provides sufficient time margin for the execution of the model-free identification of the present invention, and ensures the reliable implementation of the control strategy at high frequency.

[0033] (2) The present invention uses a data-driven autoregressive moving average model to identify system dynamics online, eliminating the dependence on precise prior physical models. This method is adaptive to dynamic changes caused by uncertainties such as motor parameter mismatch, so that the control system can still maintain high-precision current tracking and high-quality control performance when key parameters change significantly.

[0034] (3) To address the issues of missing output data and dimension mismatch in dual-rate modelless control, this invention employs an interpolation time-shift prediction method to effectively reconstruct the missing data between slow sampling periods, thereby achieving quasi-cycle-by-cycle control updates at high rates. This method significantly improves the current quality and dynamic response speed of the system, outperforming traditional non-cycle-by-cycle control schemes. Attached Figure Description

[0035] Figure 1 This is an overall structural diagram of a dual-rate model-free control system according to the present invention;

[0036] Figure 2 This is a detailed flowchart based on the dual-rate model-free control algorithm;

[0037] Figure 3 It is a digital implementation based on the dual-rate model-free control algorithm;

[0038] Figure 4 This is a waveform comparison diagram of the present invention under the condition of motor parameter mismatch. Detailed Implementation

[0039] The specific embodiments of the present invention will now be described with reference to the accompanying drawings to enable those skilled in the art to better understand the invention. It should be particularly noted that in the following description, detailed descriptions of known functions and designs that might obscure the main content of the invention will be omitted here.

[0040] Example

[0041] Figure 1 This is an overall structural diagram of a dual-rate model-free control system according to the present invention.

[0042] In this embodiment, we will first give a brief introduction to the overall structure of the dual-rate model-free control system, such as... Figure 1 As shown, it includes: a three-phase permanent magnet synchronous motor (PMSM) 1, a silicon carbide (SiC) based high-frequency driver 2, a DC power supply 3, a position encoding module 4, a current sampling module 5, a speed calculation module 6, a difference operation unit 7, and a speed loop PI controller 8. / The system includes a coordinate transformation module 9, a space vector pulse width modulation module 10, and a dual-rate model-free controller 11.

[0043] The three-phase permanent magnet synchronous motor 1 is the system control object; in this embodiment, it is a surface-mounted permanent magnet synchronous motor, and the energy input comes from the three-phase AC power of the high-frequency driver 2; the rotational position of its rotor is determined by the position encoder module 5.

[0044] The high-frequency motor driver 2 is a three-phase voltage source inverter (VSI), with its input side connected to a DC power supply 3 and its output side connected to a three-phase permanent magnet synchronous motor 1.

[0045] The DC power supply 3 is the DC voltage source of the system, providing a stable DC voltage for the high-frequency driver 2;

[0046] The position encoding module 4 is a position detection device, which is installed on the shaft of the three-phase permanent magnet synchronous motor 1 and is used to detect electrical angles. And pass it to the speed calculation module 6;

[0047] The current sampling module 5 is a current detection device, which is installed between the driver 2 and the permanent magnet synchronous motor 1 to collect three-phase AC current. And pass it on Coordinate transformation module 9;

[0048] The speed calculation module 6 is connected to the position encoding module 4, and calculates the speed based on the electrical angle. The angular velocity of the motor is calculated in real time. ;

[0049] The difference operation unit 7 is an error calculation unit, and its input comes from the reference angular velocity. Angular velocity of velocity calculation module 6 The speed error is obtained by calculating the difference between the two.

[0050] The input to the speed loop PI controller 8 is the speed error of the difference calculation unit 7, which is obtained through the proportional-integral (PI) controller. Shaft reference current Secondly, Shaft reference current In this embodiment, it is set to zero;

[0051] The / Coordinate transformation module 9 is a spatial coordinate transformation matrix, which transforms the three-phase currents respectively. Convert to Current in stationary coordinate system general shaft reference current converted to Reference current in stationary coordinate system It provides the motor current signal for the dual-rate modelless controller 11;

[0052] The input of the space vector pulse width modulation module 10 is connected to the dual-rate model-free controller 11; this module outputs the obtained high-frequency control sequence. The signals are modulated sequentially to generate the switching signals required to drive the high-frequency driver 2. ;

[0053] The dual-rate model-free controller 11 is the core of this invention; it receives data from... / The current signal from the coordinate transformation module 9 is processed by an internal algorithm and outputs a high-frequency control sequence to the space vector pulse width modulation module 10; the dual-rate modelless controller 11 includes: a time delay compensation module 12, a model coefficient identification module 13, an ARMAX driving model module 14, and an interpolation time shift prediction module 15.

[0054] The delay compensation module 12 receives The current output by coordinate transformation module 9 at the current time This module is used for online estimation of system lumped disturbances and compensation for computational delays, predicting the current in the next slow-rate cycle. And output it to the model coefficient identification module 13;

[0055] The input terminal of the model coefficient identification module 13 is connected to the time delay compensation module 12 to receive the predicted current. This module employs the Recursive Least Squares (RLS) algorithm to transform the online identified ARMAX model coefficient vectors... And output to ARMAX driver model module 14;

[0056] The input terminal of the ARMAX driving model module 14 is connected to the model coefficient identification module 13, and receives the coefficient vector updated in real time. This module is used to construct a data-driven mathematical model that characterizes the system dynamics and is called by the interpolation time-shift prediction module 15.

[0057] The interpolation time-shift prediction module 15 uses the driving model constructed by the ARMAX driving model module 14 to perform time-shift prediction; this module reconstructs the missing data through interpolation and combines it with a time-shifting strategy to obtain the optimal control sequence. The signal is then output to the space vector pulse width modulation module 10 to obtain a high-frequency switching signal.

[0058] Figure 2 The detailed flowchart of the present invention based on the dual-rate model-free control algorithm includes the following steps:

[0059] (1) Sampling slow-rate data information;

[0060] (1.1) Obtain the three-phase current of the motor at the current moment using the current sampling module. , , ;

[0061] (1.2), use The coordinate transformation module converts the three-phase current Convert to Current in stationary coordinate system ,That The coordinate transformation formula is expressed as:

[0062] ;

[0063] (1.3) Obtain the electrical angle of the motor at the current moment using the position encoding module. ;

[0064] (1.4) Differentiate the electrical angle using the velocity calculation module. Converted to electric angular velocity ;

[0065] (1.5) By referring to the given speed The electrical angular velocity error is calculated using the difference operation unit and then input into the speed loop PI controller to obtain... Shaft reference current To achieve maximum torque-to-current ratio control, Shaft reference current Set to 0;

[0066] (1.6), use The coordinate transformation module will The reference current of the rotating coordinate system is transformed into The reference current in the stationary coordinate system, its The coordinate transformation formula is:

[0067] ;

[0068] (2) Observer-based time delay compensation;

[0069] (2.1) The calculation delay and filtered noise data are compensated by the extended state observer of the hyperlocal model. The specific formula is as follows:

[0070] ;

[0071] in, For the system at slow speed The time estimation error; This is the system frequency boost factor, which can also be expressed as the slow sampling period. and control of rapid cycles The ratio is set to 4 in this embodiment; for shaft current exist The estimated value at time, ; For estimating unknowns that include disturbances, This is the current input voltage; and For feedback gain, For the input gain, it is set to 8000 in this embodiment. and 400;

[0072] (2.2) The next slow rate is obtained by prediction. Current at any moment And it serves as the starting point for derivation of the control input sequence, reducing the impact of computational delay;

[0073] (3) Data-driven model construction and identification;

[0074] (3.1) On a slow timescale Next, we construct the ARMAX-driven model, and the discretized model is represented as follows:

[0075] ;

[0076] in, , , They represent At this moment Shaft output current, input voltage, and white noise term; , and For a stable polynomial, it includes a shift operator. ,Right now Specifically, it is expressed as follows:

[0077] ;

[0078] in, , and In this embodiment, the order of the corresponding polynomial is set to 4. , and The coefficients that need to be determined in real time for the corresponding polynomial;

[0079] (3.2) To facilitate the identification of coefficients in the model, the ARMAX model is rewritten in the following form:

[0080] ;

[0081] in, and These are represented as information vector and coefficient vector, respectively, as follows:

[0082] ;

[0083] Due to information vector Only includes slow time scales The input and output data, therefore the coefficient vector Online identification is possible during slow periods. In this embodiment, the model coefficients are estimated using the Recursive Least Squares (RLS) algorithm, as shown below:

[0084] ;

[0085] in, , and These are the process error term, the gain matrix, and the covariance matrix, respectively. The forgetting factor is set to 1 in this embodiment, and the coefficients in the model can be estimated in real time using the recursive least squares algorithm.

[0086] (4) Construct a rapid rate control rate;

[0087] Considering that the motor parameters change slowly within a sufficiently small sampling interval, the slow period The internal model coefficients remain unchanged, i.e. Based on time-shift prediction strategies, on fast time scales The control rate under these conditions can be expressed as:

[0088] ;

[0089] in, ,Include The fast rate control sequence of a control input can be represented as:

[0090] ;

[0091] (5) Reconstruction of missing data and solving for the control sequence;

[0092] (5.1) When At that time, there is missing future data in the fast rate control sequence formula. and and previously missing data and In this embodiment, intermediate missing data is obtained through linear interpolation, while a deadbeat control concept is employed to ensure that the current in the next sampling cycle tracks the current reference current. The corresponding interpolation estimation equation is:

[0093] ;

[0094] in, Estimate vectors for missing data. and These are the starting and ending vectors of the known data, respectively; because... It is white noise with an expected value of 0, so it can be approximated as such. ;

[0095] (5.2) Complete the sampling interval by interpolation equation The missing data is used to substitute the reconstructed current vector into the fast rate control sequence equation to obtain a set of usable time series voltage control inputs.

[0096] (6) High-frequency drive execution;

[0097] The obtained control sequence Space vector pulse modulation is performed sequentially, and the carrier frequency and the fast rate control frequency are set to be consistent. The resulting switching signal drives the high-frequency silicon carbide motor driver.

[0098] like Figure 3 This paper demonstrates the specific implementation of the control algorithm of this invention on a digital controller. The control is decoupled into fast-rate and slow-rate interrupts through nested interrupts. The slow-rate interrupt handles all computationally intensive tasks, including stator current sampling, delay compensation, online ARMAX model identification, and recursive calculation of sequential control inputs. The fast-rate interrupt only performs SVPWM on pre-calculated control inputs. The nested interrupt implementation designed in this invention effectively achieves quasi-periodic control performance, realizes control updates within each fast cycle, and avoids the computational burden of high-frequency interrupts. The dual-rate framework differs from the traditional single-rate implementation because it requires an additional carrier frequency to match the control frequency, thereby accommodating multiple modulation signals generated in each slow-rate interrupt.

[0099] This embodiment will be described with reference to examples, such as... Figure 4 The diagram shows experimental waveforms comparing the simultaneous mismatch of motor resistance, inductance, and flux linkage parameters. The sampling and switching frequencies for both schemes were set to 20 kHz and 80 kHz, respectively. The controller values ​​were set to 10% of the actual resistance, 20% of the actual inductance, and 150% of the actual flux linkage. From top to bottom, the diagrams represent motor speeds. electromagnetic torque Phase current The parameter mismatch trigger flag is shown in (a) for the traditional model-based dual-rate control scheme and (b) for the control scheme proposed in this invention. It can be seen that the total harmonic distortion of the current in the traditional model-based dual-rate control scheme is as high as 7.15% under parameter mismatch, and the speed and electromagnetic torque waveforms fluctuate significantly after parameter mismatch switching. However, the total harmonic distortion of the current in the control scheme proposed in this invention is significantly reduced to 2.92%, showing better current quality and demonstrating the advantages of the dual-rate model-free control scheme.

[0100] Although the illustrative specific embodiments of the present invention have been described above to enable those skilled in the art to understand the invention, it should be understood that the invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the invention as defined and determined by the appended claims, and all inventions utilizing the concept of the present invention are protected.

Claims

1. A dual-rate model-free control system for high-frequency motor drive, characterized in that, include: Three-phase permanent magnet synchronous motor (PMSM), silicon carbide-based high-frequency driver, DC power supply, position encoding module, current sampling module, speed calculation module, difference calculation unit, speed loop PI controller. / Coordinate transformation module, space vector pulse width modulation module, and dual-rate model-free controller; The dual-rate model-free controller internally includes: a time delay compensation module, a model coefficient identification module, an ARMAX driving model module, and an interpolation time shift prediction module; The delay compensation module receives Current output by the coordinate transformation module at the current time This module is used for online estimation of system lumped disturbances and compensation for computational delays, predicting the current in the next slow-rate cycle. And output it to the model coefficient identification module; The input terminal of the model coefficient identification module is connected to the time delay compensation module to receive the predicted current. This module uses the Recursive Least Squares (RLS) algorithm to identify the coefficient vectors of the ARMAX driving model online. And output it to the ARMAX driving model module; The input terminal of the ARMAX-driven model module is connected to the model coefficient identification module, and receives the coefficient vector updated in real time. This module is used to construct a data-driven mathematical model that characterizes the system dynamics and is called by the interpolation time-shift prediction module. The interpolation time-shift prediction module uses the driving model constructed by the ARMAX driving model module to perform time-shift prediction; this module reconstructs missing data through interpolation and combines it with a time-shifting strategy to obtain the optimal control sequence. After being output to the space vector pulse width modulation module, a high-frequency switching signal is obtained; The dual-rate model-free controller receives from / The current signal from the coordinate transformation module is processed by an internal algorithm and then outputs a high-frequency control sequence to the space vector pulse width modulation module. The input of the space vector pulse width modulation module is connected to a dual-rate model-free controller, used to process the obtained high-frequency control sequence. The signals are modulated sequentially to generate the high-frequency switching signals required to drive the high-frequency driver. .

2. The dual-rate model-free control system for high-frequency motor drive according to claim 1, characterized in that, The three-phase permanent magnet synchronous motor is the system control object, and its energy input comes from the three-phase AC power of the high-frequency driver; The high-frequency driver is a three-phase voltage source inverter (VSI), with its input side connected to a DC power supply and its output side connected to a three-phase permanent magnet synchronous motor. The DC power supply is the DC voltage source of the system, providing a stable DC voltage for the high-frequency driver; The position encoding module is a position detection device, which is installed on the shaft of the three-phase permanent magnet synchronous motor and is used to detect electrical angles. And pass it to the speed calculation module; The current sampling module is a current detection device installed between the high-frequency driver and the permanent magnet synchronous motor, used to collect three-phase AC current. And pass it on Coordinate transformation module; The velocity calculation module is connected to the position encoding module, and calculates the velocity based on the electrical angle. The angular velocity of the motor is calculated in real time. ; The difference operation unit is an error calculation unit, and its input comes from the reference angular velocity. Angular velocity of the velocity calculation module The speed error is obtained by calculating the difference between the two. The input to the speed loop PI controller is the speed error of the difference calculation unit, which is obtained through the proportional-integral controller. Shaft reference current ; The / The coordinate transformation module is a spatial coordinate transformation matrix, which transforms the three-phase currents respectively. Convert to Current in stationary coordinate system general shaft reference current converted to Reference current in stationary coordinate system .

3. A dual-rate model-free control method utilizing the dual-rate model-free control system of claim 1, characterized in that, Includes the following steps: (1) Sampling slow-rate data information; (1.1) Obtain the three-phase current of the motor at the current moment using the current sampling module. , , ; (1.2), use The coordinate transformation module converts the three-phase current Convert to Current in stationary coordinate system ; (1.3) Obtain the electrical angle of the motor at the current moment using the position encoding module. ; (1.4) Differentiate the electrical angle using the velocity calculation module. Converted to electric angular velocity ; (1.5) By referring to the given speed The electrical angular velocity error is calculated using the difference operation unit and then input into the speed loop PI controller to obtain... Shaft reference current In addition, Shaft reference current Set to 0; (1.6), use The coordinate transformation module will The reference current of the rotating coordinate system is transformed into Reference current in stationary coordinate system , ; (2) Observer-based time delay compensation; (2.1) The calculation delay and filtered noise data are compensated by the extended state observer of the hyperlocal model. The specific formula is as follows: ； in, For the system at slow speed The time estimation error; This represents the system frequency boost factor, and also indicates the slow sampling period. and control of rapid cycles The ratio; for shaft current exist The estimated value at time, ; For estimating unknowns that include disturbances, This is the current input voltage; and For feedback gain, Input gain; (2.2) The next slow rate is obtained by prediction. Current at any moment And serve as the starting point for derivation of the control input sequence; (3) Data-driven model construction and identification; (3.1) On a slow timescale Next, we construct the ARMAX-driven model, whose discretized model representation is as follows: ； in, , , They represent At this moment The shaft's output current, input voltage, and white noise term; , and For a stable polynomial, it includes a shift operator. ,Right now Specifically, it is expressed as follows: ； in, , and For the order of the corresponding polynomial, , and The coefficients that need to be determined in real time for the corresponding polynomial; (3.2) To facilitate the identification of coefficients in the model, the ARMAX model is rewritten in the following form: ； in, and These are represented as information vector and coefficient vector, respectively, as follows: ； The model coefficients are estimated using the recursive least squares algorithm, as follows: ； in, , and These are the process error term, gain matrix, and covariance matrix, respectively. As a forgetting factor, the coefficients in the model can be estimated in real time using a recursive least squares algorithm; (4) Construct a rapid rate control rate; Considering that the motor parameters change slowly within a sufficiently small sampling interval, the slow period The internal model coefficients remain unchanged, i.e. Based on time-shift prediction strategies, on fast time scales The control rate is expressed as follows: ； in, ,Include The fast rate control sequence of each control input is represented as: ； (5) Reconstruction of missing data and solving for the control sequence; (5.1) Obtain future missing data in the control sequence expression by linear interpolation. and and previously missing data and And ensure that the current in the next sampling period tracks the current reference current. Its interpolation estimation equation is: ； in, Estimate vectors for missing data. and These are the starting and ending vectors of the known data, respectively; because... It is white noise with an expected value of 0, so it can be approximated as such. ; (5.2) Complete the sampling interval by interpolation equation The missing data is used to substitute the reconstructed current vector into the fast rate control sequence equation to obtain a set of usable time series voltage control inputs. (6) High-frequency drive execution; The obtained control sequence Space vector pulse modulation is performed sequentially, and the carrier frequency and the fast rate control frequency are set to be consistent. The resulting switching signal drives the high-frequency silicon carbide motor driver.

Images