Transfer function identification method based on iterative optimization

Abstract

The present invention relates to the technical field of motor control, and provides a transfer function identification method based on iterative optimization. The transfer function identification method comprises the following steps: model derivation; initial value setting; actual data measurement; iterative optimization; stable output; and convergence determination. According to the present invention, an iterative learning filter is integrated into a batch least squares identification algorithm, so that the identified data is more accurate, the design of a high-accuracy controller is facilitated, and the anti-interference ability is strong, thereby achieving high parameter identification accuracy, and achieving high-speed and high-accuracy control of a system. An approximately feasible optimal solution is obtained by calculating by means of torque estimation, so that the reliability of the solution is ensured and the solution can better approximate an accurate value compared to a manually estimated initial value. Not only the step of manually estimating the initial value can be effectively eliminated, but also the identification accuracy of the transfer function is ensured, improving the open-loop precision and response speed of an engine.

Description

A transfer function identification method based on iterative optimization Technical Field

[0001] The present invention relates to the technical field of motor control, and in particular to a transfer function identification method based on iterative optimization. Background Art

[0002] With the deepening of control theory research and the development of artificial intelligence technology, the use of advanced technologies to simplify complex technologies, implement previously unattainable functional improvements, or improve control accuracy are all hot research areas. Precise plant models make it possible to use control models to correct control outputs, while also offsetting steady-state fluctuations caused by closed-loop regulation, achieving better control results.

[0003] To achieve carbon neutrality, we have entered the era of electric vehicles. To meet the higher demands for electric motor energy conservation and improved performance, we need to design higher-performance motor control systems. However, determining a reasonable and simplified motor control model is crucial for analyzing and designing motor control systems. To ensure smoother and more reliable motor operation, we must know the precise model parameters of this system.

[0004] Currently, most identification methods based on the CARMA model (autoregressive moving average model with control variables) perform filtering and identification separately, and the filter parameters are all fixed values. This results in inflexibility and poor adaptability of the identification algorithm, and it cannot effectively filter out environmental interference and measurement noise that exist in actual applications, which in turn affects the accuracy of the identification parameters. However, solving the optimal value through theoretical calculation of nonlinear least squares method and optimizing the initial value by torque estimation can shorten the transfer function solution time and improve the identification accuracy. However, the current least squares-based method for identifying the motor open-loop control transfer function requires manual estimation of the initial value. Moreover, without a suitable filter, the identification result will lead to large errors in the motor transfer function solution, affecting the motor control accuracy.

[0005] Summary of the Invention

[0006] In view of the above problems, the present invention provides a transfer function identification method based on iterative optimization, aiming to solve the technical problems raised in the above background technology, such as the need for manual estimation of initial values, weak anti-interference ability, and low parameter identification accuracy.

[0007] To achieve the above object, the present invention provides the following technical solution: a transfer function identification method based on iterative optimization, comprising the following steps:

[0008] S1. Model Derivation: Based on the electrical and mechanical structures of the motor control system, derive the corresponding transfer function model and introduce multiple biquadratic links into the transfer function.

[0009] S2. Set initial values: Solve the transfer function of the target model and set the initial values ​​of the undetermined parameters of the reference models of multiple second-order transfer functions to 1;

[0010] S3. Measured Data: Install the target engine on a torque tester, measure the relationship between the target torque and the calibrated torque, determine the proportionality between the actual torque and the predicted torque, and modify the transfer function parameters. This data serves as a reference for model parameter identification.

[0011] S4. Iterative Optimization: Based on the transfer function model derived in step S2, determine the system parameters that need to be identified. First, use the torque estimation method to calculate an approximately feasible optimization solution. Then, use this optimization solution as the initial value of the nonlinear least squares method, and further iteratively optimize to obtain the optimal solution.

[0012] S5. Stable output: The output value obtained in step S3 is filtered by a filter, and the filtered output value is returned to step S3;

[0013] S6. Convergence determination: If the identification result meets the convergence requirement, the identification is completed; otherwise, return to step S3. Further, the transfer function model derived in step S1 is as follows:

[0014] Furthermore, in step S3, the torque estimation proportional coefficients corresponding to different working conditions are different, and the torque test needs to be performed under at least four working conditions: starting, idling, sudden acceleration from idling, and 100% load.

[0015] Furthermore, in order to make the identification more accurate in step S3, at least four linear input forms are required to perform torque testing on the same target model.

[0016] Furthermore, the four linear input forms are: x = tx = t / 2 x = log a (t+1),a>1 x=a t ,a>1

[0017] x is the input value, t is the time, and a is a constant.

[0018] Furthermore, in step S3, the torque proportional coefficient is estimated by setting the target torque using the motor calibration system on the controller side, calculating the calibrated torque in real time, and inferring the internal torque consumption of the motor based on the difference between the target torque and the calibrated torque. The torque consumption curve is fitted through the internal consumption of the motor under different working conditions, and the predicted torque is deduced in combination with the calibrated torque. The torque output of the dynamometer is measured, and the torque estimation proportional coefficient table is solved according to the ratio between the predicted torque and the actual torque output.

[0019] Furthermore, the model with the determined order is transformed from the s domain to the z domain formula through the bilinear transformation formula;

[0020] The bilinear transformation formula is:

[0021] The z-domain formula is:

[0022] Where G(s) is the transfer function, T is the sampling period, z is the delay operator, α0…α n-1 are the coefficients that need to be identified in the numerator of the discretized transfer function, β1…β n The coefficients that need to be identified for the denominator of the discretized transfer function;

[0023] From β1…β n composition Filter, set when filtering the output data for the first time

[0024] Output data w The data after filtering is

[0025] Furthermore, in step S5, the simulation calculation after the transfer function identification must be verified in closed-loop control.

[0026] Furthermore, in step S5, the error analysis of the identification accuracy is to analyze the error between the transfer function simulation control quantity and the measured open-loop control measurement data of the same set of control parameters.

[0027] Furthermore, before step S3, the output data is first filtered using a filter; the numerator coefficient of the filter is 1, the initial value of the denominator coefficient is 1, and each subsequent iteration updates the last identification result as the denominator coefficient.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] By integrating iterative learning filters into the batch least squares identification algorithm, the identified data is more accurate, which is conducive to the design of high-precision controllers, has strong anti-interference ability, and high parameter identification accuracy, which helps to achieve high-speed and high-precision control of the system; and through repeated identification of multiple working conditions and repeated convergence after multiple iterations of data, it is beneficial to convergence accuracy and further improves the high identification accuracy; through torque estimation calculation, an approximate feasible optimization solution is obtained, which can ensure the reliability of the solution and can better approximate the accurate value compared with the manually estimated initial value; it can not only effectively eliminate the step of manually estimating the initial value, but also ensure the identification accuracy of the transfer function, which is conducive to improving the open-loop accuracy and response speed of the engine. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] FIG1 is a schematic flow chart of the method of the present invention; DETAILED DESCRIPTION

[0031] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be understood as limiting the present invention.

[0032] The following describes an embodiment of the present invention based on its overall structure.

[0033] In the embodiment, please refer to FIG1 , a transfer function identification method based on iterative optimization includes the following steps:

[0034] S1. Model Derivation: Based on the electrical and mechanical structures of the motor control system, derive the corresponding transfer function model and introduce multiple biquadratic links into the transfer function.

[0035] S2. Set initial values: Solve the transfer function of the target model and set the initial values ​​of the undetermined parameters of the reference models of multiple second-order transfer functions to 1;

[0036] S3. Measured Data: Install the target engine on a torque tester, measure the relationship between the target torque and the calibrated torque, determine the proportionality between the actual torque and the predicted torque, and modify the transfer function parameters. This data serves as a reference for model parameter identification.

[0037] S4. Iterative Optimization: Based on the transfer function model derived in step S2, determine the system parameters that need to be identified. First, use the torque estimation method to calculate an approximately feasible optimization solution. Then, use this optimization solution as the initial value of the nonlinear least squares method, and further iteratively optimize to obtain the optimal solution.

[0038] S5. Stable output: The output value obtained in step S3 is filtered by a filter, and the filtered output value is returned to step S3;

[0039] S6. Convergence determination: If the identification result meets the convergence requirement, the identification is completed; otherwise, return to step S3.

[0040] Although an embodiment of the present invention has been shown and described, this specific embodiment is merely an explanation of the present invention and is not a limitation of the invention. The specific features, structures, materials or characteristics described may be combined in an appropriate manner in any one or more embodiments or examples. After reading this specification, those skilled in the art may make modifications, substitutions and variations to the embodiment without creative contribution as needed without departing from the principles and purpose of the present invention. However, as long as they are within the scope of the claims of the present invention, they are protected by patent law.

Claims

1. A transfer function identification method based on iterative optimization, characterized in that: The steps include: S1. Derivation model: According to the electrical structure and mechanical structure of the motor control system, the corresponding transfer function model of operating power and output torque is derived, and multiple double second-order links are introduced into the transfer function; S2. Set initial values: solve the transfer function of the target model and set the initial values ​​of the undetermined parameters of the reference models of multiple second-order transfer functions to 1; S3. Measured data: Install the target model on the torque test device, test the relationship between the target torque and the calibrated torque, determine the proportionality coefficient between the actual torque and the predicted torque, and correct the transfer function parameters, which will be used as a reference object for model parameter identification; S4. Iterative optimization: According to the transfer function model derived in step S2, the system parameters that need to be identified are determined, and a torque estimation method is first used to calculate an approximately feasible optimization solution, and then the optimization solution is used as the initial value of the nonlinear least squares method, and further iterative optimization is performed to obtain the optimal solution; S5. Stable output: filtering the actual torque value obtained in step S3 through a filter, and returning the filtered output value to step S3; S6. Convergence determination: if the identification result meets the convergence requirement, the identification is completed; otherwise, return to step S3.

2. The transfer function identification method based on iterative optimization according to claim 1, characterized in that: The transfer function model derived in step S1 is as follows: Among them, k1, k2, k3, k4, k5, k6, k7, and k8 are unknown parameters that need to be fitted in the transfer function of the motor control system model.

3. The transfer function identification method based on iterative optimization according to claim 1, characterized in that: In step S3, the torque estimation proportional coefficients corresponding to different working conditions are different. For the same target model, the torque test needs to be performed under at least four working modes: starting, idling, sudden acceleration at idling, and 100% load.

4. The transfer function identification method based on iterative optimization according to claim 1, characterized in that: In order to make the identification more accurate in step S3, at least four input forms are required to perform torque testing on the same target model.

5. The transfer function identification method based on iterative optimization according to claim 4, characterized in that: The four input forms are: x = tx = t / 2 x = log a (t+1),a>1 x=a t ,a>1 x is the input value, t is the time, and a is a constant.

6. The transfer function identification method based on iterative optimization according to claim 1, characterized in that: In step S3, the torque proportional coefficient is estimated by setting the target torque with the motor calibration system on the controller side, calculating the calibrated torque in real time, and calculating the internal torque consumption of the motor according to the difference between the target torque and the calibrated torque. The torque consumption curve is fitted by the internal consumption of the motor under different working conditions, and the predicted torque is deduced in combination with the calibrated torque. The torque output of the dynamometer is measured, and the torque estimation proportional coefficient table is solved according to the ratio between the predicted torque and the actual torque output.

7. The transfer function identification method based on iterative optimization according to claim 1, characterized in that: In step S5, the simulation calculation after the transfer function identification must be verified in the closed-loop control.

8. The transfer function identification method based on iterative optimization according to claim 1, characterized in that: In step S5, the error analysis of identification accuracy is to analyze the error between the transfer function simulation control quantity and the measured open-loop control measurement data of the same set of control parameters.

9. The transfer function identification method based on iterative optimization according to claim 1, characterized in that: Before step S3, the output data is first filtered by a filter; the numerator coefficient of the filter is 1, and the initial value of the denominator coefficient is 1, and then each iteration updates the last identification result as the denominator coefficient.

Images