CN-122001258-A - Motor control system based on self-learning gain

CN122001258ACN 122001258 ACN122001258 ACN 122001258ACN-122001258-A

Abstract

The invention discloses a motor control system based on self-learning gain, which comprises a self-learning gain supercoiled sliding mode observer, wherein the sliding mode gain is dynamically adjusted by monitoring current observation errors in real time, so that the sliding mode gain is rapidly increased to enhance robustness when the system is disturbed, and is automatically reduced to inhibit buffeting when the system tends to be stable, thereby realizing rotor position and rotating speed estimation with high precision and strong robustness.

Inventors

HU DANFENG
Tu Feiyu
YUAN ZIJING
CAO HENGWEI

Assignees

苏州大学

Dates

Publication Date: 20260508
Application Date: 20260108

Claims (6)

1. The motor control system based on the self-learning gain is characterized by comprising a supercoiled sliding mode observer, wherein the calculation equation of the supercoiled sliding mode observer is as follows: Wherein, the , , 、 The self-adaptive learning gain update law of the sliding mode gain is as follows: Wherein, the In order to observe the error in the current, For the margin of error to be preset, Is a positive constant.
2. The self-learning gain based motor control system of claim 1 wherein the supercoiled sliding mode observer is configured to input motor stator voltage And current And estimating the counter electromotive force component of the stator 、。
3. The self-learning gain based motor control system of claim 1, wherein when At the time, the sliding mode gain Start to increase when At the time, the sliding mode gain The decrease starts.
4. The self-learning gain based motor control system of claim 1 wherein the sliding mode gain 、 The following relationship is followed: 。
5. The self-learning gain based motor control system of claim 1, further comprising: A speed adjusting ring for inputting a given rotation speed of the motor control system Correlating it with the observed motor speed Comparing, forming a speed error, outputting a torque current set through the PI regulator ; Current regulation loop for supplying three-phase current 、、 Obtaining a two-phase static coordinate system current through Clark transformation 、 Then the Park is transformed into a rotating coordinate system to obtain 、 Will be 、 Respectively and practically match 、 Comparing, and obtaining d-axis and q-axis voltage instructions after two paths of PI regulation 、 ; A space vector pulse width modulation module for inputting voltage command obtained by inverse Park conversion 、 Outputting drive signals of three-phase bridge arms Controlling the inverter to output corresponding three-phase voltages; A phase-locked loop for phase locking the back electromotive force vector to obtain a motor rotor position estimation value And velocity estimation ; Inverter for generating DC bus voltage Under the power supply, according to the driving signal Outputting three-phase alternating voltage to motor The motor operates to produce three-phase current And feeding back to the motor control system.
6. The self-learning gain based motor control system of claim 5 wherein the direct-axis current setting is given as The FOC control of the motor is realized.

Description

Motor control system based on self-learning gain Technical Field The invention relates to the technical field of permanent magnet synchronous motor control, in particular to a motor control system based on self-learning gain. Background Permanent magnet synchronous motors (PERMANENT MAGNET Synchronous Motor, PMSM) are widely used in the fields of industrial servo, electric vehicles, aerospace and the like due to high efficiency, high power density and excellent control performance. To achieve high performance vector control, it is often necessary to install position sensors (e.g., encoders) to obtain rotor position and rotational speed information in real time. However, the introduction of the position sensor not only increases the cost and volume of the system, but also reduces the reliability and robustness of the system in severe environments such as high temperature, vibration, and the like. Therefore, sensorless control technology that does not require a physical sensor is an important research direction in the field of PMSM control. At present, the PMSM sensorless control method is mainly divided into two types, namely a high-frequency signal injection method based on salient pole effect and a model method based on motor back electromotive force. The high frequency signal injection method is suitable for the zero low speed region, but introduces additional noise and harmonic interference. The model method is suitable for medium-high speed areas, wherein the sliding mode observer (Sliding Mode Observer, SMO) has strong robustness to parameter perturbation and external disturbance, and has the advantages of simple structure, easy realization and wide attention. However, the conventional sliding mode observer introduces serious high-frequency buffeting due to discontinuous sign functions, which affects the control accuracy of the system and even leads to instability of the system. To suppress buffeting, a supercoiled sliding Mode Observer (Super-TWISTING SLIDING Mode Observer, STSMO) is proposed. The observer converts discontinuous control law into continuous control law by introducing integral term, effectively smoothes control signals, remarkably reduces buffeting phenomenon and improves observation performance. Conventional STSMO typically employs a fixed gain or relies solely on motor speed for gain adjustment. The method based on the fixed gain is difficult to maintain optimal performance in a full-speed domain and under different working conditions, namely, the slow convergence speed and the slow dynamic response of an observer are caused by the excessively low gain, and buffeting is reintroduced in a steady state when the gain is excessively high, so that estimation accuracy is influenced. The gain adjustment strategy which only depends on the rotation speed has too small gain at low speed, which may cause the observer to be unable to start normally, and the gain may be too high at high speed, which aggravates the noise amplification problem. In the prior art, an adaptive gain strategy combining rotation speed and current errors is proposed, but the methods either depend on accurate disturbance upper bound information or have complex switching logic, so that stable and reliable application is difficult to realize in engineering. In summary, the existing sensor-free control method based on STSMO still has the defects in the aspect of gain self-adaptive adjustment, so that the buffeting is difficult to be effectively inhibited while the dynamic response speed is ensured, and particularly, the observation performance is easy to be obviously reduced when the motor parameter mismatch (such as stator resistance change) and the large-range abrupt change of the rotating speed are faced. Therefore, how to design a self-adaptive gain strategy which can adapt to different working conditions, does not need to accurately disturb the upper bound and is simple to realize, so that buffeting is effectively inhibited while dynamic response speed is ensured, and the self-adaptive gain strategy is a problem to be solved by a person skilled in the art. Disclosure of Invention The invention aims to provide a motor control system based on self-learning gain, which comprises a self-learning gain supercoiled sliding mode observer, wherein the sliding mode gain is dynamically adjusted by monitoring current observation errors in real time, so that the sliding mode gain is rapidly increased to enhance robustness when the system is disturbed, and is automatically reduced to inhibit buffeting when the system tends to be stable, thereby realizing rotor position and rotating speed estimation with high precision and strong robustness. The technical scheme of the invention is that the motor control system based on self-learning gain comprises a supercoiled sliding mode observer, wherein the operation equation of the supercoiled sliding mode observer is as follows: Wherein, the ,,、The self-adaptive learning gain update law of the