Neural network method-based rotor position compensation method without position sensor

Abstract

The invention relates to a position sensorless rotor position compensation method based on a neural network method, which is based on a neural network learning prediction model and comprises the following steps: firstly, acquiring three-phase current, stator resistance, inductance information, motor temperature, rotating speed and torque of a motor; calculating the back electromotive force of the motor according to the collected data, and calculating the rotor position estimation information by using the back electromotive force; and inputting rotor position estimation information, motor rotor temperature, rotating speed, torque and other information into the neural network learning prediction model, calculating a rotor position estimation value and a compensation value to obtain a rotor position prediction value, and using the rotor position prediction value for the rotor position in the next carrier frequency period. According to the method, the rotor position precision is high, high-dynamic control of the motor can be met, meanwhile, self-adaptive change can be achieved according to different rotating speeds, torques and temperatures, and the rotor position requirements of the motor under various conditions are met; in addition, the neural network rotor position estimation model establishment method is simple and reliable.

Description

Technical Field

[0001] The invention belongs to the technical field of permanent magnet synchronous motors, and in particular relates to a position sensorless rotor position compensation method based on a neural network method. Background Art

[0002] With the increase of output power and power density of permanent magnet synchronous motors, higher requirements are placed on the control accuracy and efficiency of motors. However, traditional motor position sensors, such as resolvers, will affect the stability of motor torque control output due to manufacturing accuracy, assembly errors, and shortcomings in their own design characteristics. At present, position sensorless control methods are gradually coming into people's view. Position sensorless control methods not only save motor position sensors, making the motor structure more compact, but also make the motor more reliable, and the control accuracy and control efficiency can be significantly improved.

[0003] If a rotor position compensation method can be developed, the rotor position compensation value can be obtained through the rotor position estimation information, the rotor position estimation value and the compensation value can be calculated to obtain the rotor position prediction value, and this value can be used for the rotor position of the next carrier frequency cycle. This is very necessary and will definitely have broad application prospects. Summary of the invention

[0004] The purpose of the present invention is to provide a position sensorless rotor position compensation method based on a neural network method to solve the problem of obtaining high-precision rotor position, satisfying high dynamic control of the motor, and being able to adaptively change according to different speed, torque and temperature.

[0005] The objective of the present invention is achieved through the following technical solutions:

[0006] A position sensorless rotor position compensation method based on a neural network method is based on a neural network learning prediction model and includes the following steps: first, collecting and acquiring the three-phase current, stator resistance, inductance information, motor temperature, speed, and torque of a motor; then calculating the back electromotive force of the motor according to the collected data, and using the back electromotive force to calculate the rotor position estimation information; then inputting the rotor position estimation information, the motor rotor temperature, speed, torque and other information into the neural network learning prediction model, calculating the rotor position estimation value and the compensation value to obtain the rotor position prediction value, and using the value for the rotor position of the next carrier frequency cycle.

[0007] Further, the following steps are included:

[0008] A. Collect the motor's three-phase current, motor temperature, stator resistance, speed and torque parameters to obtain more data for estimating the rotor position and confirming the motor's working condition;

[0009] B. Calculate an estimated value of the motor rotor position based on the motor operation information obtained in step A;

[0010] C. According to the estimated value of the rotor position and the motor temperature, speed, and torque, the above parameter information is used as input to the established neural network-based permanent magnet synchronous motor rotor position dynamic compensation module, and the motor rotor position compensation value is output;

[0011] D. Based on the rotor position compensation value and the rotor position estimation value, the rotor position value of the next carrier frequency cycle is obtained.

[0012] Furthermore, in step A, the selection of parameters needs to be based on the application platform, among which temperature, speed, and torque are operating condition confirmation parameters and are mandatory items; three-phase current, stator resistance, etc. are non-mandatory items, which are the parameters required by the rotor position estimation method selected by the application platform.

[0013] Furthermore, in step B, the type of motor information to be collected in the first stage is adjusted.

[0014] Furthermore, in step C, a dynamic compensation module for the rotor position of a permanent magnet synchronous motor based on a neural network is established, including parameter setting and parameter calibration of the neural network model.

[0015] Furthermore, the parameter setting of the neural network model includes setting the input layer, output layer and output conditions.

[0016] Furthermore, the input layer is motor rotor position estimation information, motor speed, motor torque, and motor temperature.

[0017] Furthermore, the output layer is the motor rotor position compensation value.

[0018] Furthermore, the output condition is that the difference between the rotor estimated value and the actual value is less than or equal to a rotor position compensation threshold, and the compensation threshold is a maximum value of the motor rotor position accuracy required for precise motor control.

[0019] Furthermore, the parameter calibration process includes the following steps:

[0020] a. Coaxially connect multiple standard motors and high-precision photoelectric encoders on a motor stand, collect rotor position signals and speed signals output by the photoelectric encoder, decode the position signals and speed signals and input them into the motor control system for precise motor control;

[0021] b. Collect the three-phase current, motor temperature, stator resistance, speed, and torque of the motor during the calibration process, calculate the estimated value of the rotor position and the motor operating conditions, and calculate the estimated value of the motor rotor position based on the above collected information;

[0022] c. Input the estimated value of the rotor position, the actual value of the motor position calculated by the photoelectric encoder, the speed, torque, and temperature information into the dynamic compensation module of the rotor position of the permanent magnet synchronous motor based on the neural network, sum the calculated rotor position compensation value with the estimated value, and calculate the difference with the actual value solved by the photoelectric encoder to determine whether the difference meets the expected value, place a large amount of test data in the neural network learning system, and adjust the dynamic compensation module of the rotor position of the permanent magnet synchronous motor based on the neural network to meet the motor precision control requirements in the full speed range and full torque range;

[0023] d. After the dynamic compensation module of the permanent magnet synchronous motor rotor position based on the neural network is adjusted, the control verification is carried out on the bench test to verify that its reliability meets the motor control requirements.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] The present invention proposes a rotor position compensation method based on a neural network learning prediction model. The rotor position estimation information, motor rotor temperature, speed, torque and other information are input into the neural network learning prediction model to obtain a rotor position compensation value. The rotor position estimation value and the compensation value are calculated to obtain a rotor position prediction value, which is used for the rotor position of the next carrier frequency cycle. The rotor position in the present method has high accuracy and can meet the high dynamic control of the motor. At the same time, it can adapt to changes according to the speed, torque and temperature to meet the rotor position requirements of the motor under various conditions. In addition, the neural network rotor position estimation model establishment method of the present invention is simple and reliable. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0027] Figure 1 The design flow chart of the position sensorless rotor position compensation method based on the neural network method of the present invention;

[0028] Figure 2 A flow chart of the steps of the position sensorless rotor position compensation method based on the neural network method of the present invention. DETAILED DESCRIPTION

[0029] The present invention will be further described below in conjunction with embodiments:

[0030] The present invention proposes a rotor position compensation method based on a neural network learning prediction model, such as Figure 1As shown, the following steps are included: first, the three-phase current, stator resistance, inductance information, motor temperature, speed, and torque of the motor are collected and acquired; then the back electromotive force of the motor is calculated based on the collected data, and the rotor position estimation information is calculated using the back electromotive force; then the rotor position estimation information, motor rotor temperature, speed, torque and other information are input into the neural network learning prediction model, and the rotor position estimation value and compensation value are calculated to obtain the rotor position prediction value (based on the neural network dynamic angle compensation module, the rotor position compensation value is obtained, based on the angle compensation value and the angle estimation value), and the value is used for the rotor position of the next carrier frequency cycle.

[0031] The present invention provides a position sensorless rotor position compensation method based on a neural network method, comprising the following steps:

[0032] 1. Collect the motor's three-phase current, motor temperature, stator resistance, speed, torque and other parameters. The selection of these parameters needs to be based on the application platform. Among them, temperature, speed and torque are working condition confirmation parameters and are mandatory. Three-phase current and stator resistance are optional. These parameters are required by the rotor position estimation method selected by the application platform. This stage is to obtain more data for estimating the rotor position and confirming the motor working condition.

[0033] 2. Calculate the estimated value of the motor rotor position based on the motor operation information obtained in the first stage; the motor rotor position estimation method is not unique and is not limited to a certain method. Relevant personnel can choose a suitable rotor position estimation method and adjust the type of motor information to be collected in the first stage;

[0034] 3. According to the estimated value of the rotor position and the motor temperature, speed and torque, the above parameter information is used as input to the established permanent magnet synchronous motor rotor position dynamic compensation module based on neural network, and the motor rotor position compensation value is obtained as output;

[0035] The design method of the dynamic compensation module of the permanent magnet synchronous motor rotor position based on neural network is as follows:

[0036] 1. Design of dynamic compensation module for permanent magnet synchronous motor rotor position based on neural network

[0037] The key point of the neural network-based rotor position dynamic compensation module design is the parameter selection of the neural network model. The parameter selection in the neural network design is crucial, as it is related to the accuracy of the final compensation result. The parameter confirmation condition is a parameter that is strongly related to the motor state position, and the output condition of the rotor position compensation value is also crucial. On the one hand, the higher the accuracy of the rotor position compensation value, the longer the optimization calculation time. On the other hand, when the accuracy of the rotor position compensation value is too low, it will affect the motor control accuracy and motor control stability. The selection of the neural network is not the focus of this patent. Researchers can select a suitable neural network dynamic model according to actual conditions.

[0038] The parameters of the present invention are set as follows:

[0039] The input layer is: motor rotor position estimation information, motor speed, motor torque, and motor temperature;

[0040] The output layer is: motor rotor position compensation value;

[0041] Output conditions: The difference between the estimated value and the actual value of the rotor is less than or equal to the rotor position compensation threshold, which is the maximum value of the motor rotor position accuracy required for precise motor control;

[0042] 2. Parameter calibration process of permanent magnet synchronous motor rotor position dynamic compensation module based on neural network

[0043] A plurality of standard motors are coaxially connected with a high-precision photoelectric encoder on a motor test bench, wherein the rotor position signal and the speed signal output by the photoelectric encoder are collected, and the position signal and the speed signal are decoded and input into the motor control system for precise control of the motor; the three-phase current, motor temperature, stator resistance, speed, and torque of the motor during the calibration process are collected, and the estimated value of the rotor position and the motor operating condition are calculated, and the estimated value of the motor rotor position is calculated based on the above collected information; then, the estimated value of the rotor position, the actual value of the motor position calculated by the photoelectric encoder, the speed, the torque, and the temperature information are input into a dynamic compensation module for the rotor position of a permanent magnet synchronous motor based on a neural network, the calculated rotor position compensation value is summed with the estimated value, and the difference is calculated with the actual value solved by the photoelectric encoder to determine whether the difference meets the expected value, a large amount of test data is placed in a neural network learning system, and the dynamic compensation module for the rotor position of a permanent magnet synchronous motor based on a neural network is adjusted to meet the requirements of precise control of the motor in the full speed range and the full torque range; after the adjustment of the dynamic compensation module for the rotor position of a permanent magnet synchronous motor based on a neural network is completed, control verification is carried out on a test bench to verify that its reliability meets the requirements of motor control.

[0044] 4. Based on the rotor position compensation value and the rotor position estimation value, the rotor position value of the next carrier frequency cycle is obtained.

[0045] The rotor position obtained by the neural network-based rotor position compensation method proposed in the present invention has high accuracy and can meet the high dynamic control of the motor. At the same time, it can adapt to changes according to the differences in speed, torque and temperature, and can meet the rotor position requirements of the motor under various conditions. The neural network-based rotor position compensation method proposed in the present invention has wide applicability and can be used in different motor control methods. The neural network-based rotor position compensation method proposed in the present invention requires a large number of experiments and calculations in the early stage of learning and adjusting the neural network permanent magnet synchronous motor rotor position dynamic compensation module, but after the model is successfully adjusted, the module has the ability to self-adjust and learn, saving resources and time in the later stage of development.

[0046] Example 1

[0047] A method for compensating a rotor position without a position sensor based on a neural network method comprises the following steps:

[0048] 1. Collect the motor's three-phase current, motor temperature, stator resistance, speed, and torque parameters to obtain more data for estimating the rotor position and confirming the motor's working condition. The selection of parameters must be based on the application platform, among which temperature, speed, and torque are working condition confirmation parameters and are mandatory; three-phase current, stator resistance, etc. are optional and are required by the rotor position estimation method selected by the application platform.

[0049] 2. Calculate the estimated value of the motor rotor position based on the motor operation information obtained in step 1. Adjust the type of motor information to be collected in the first stage.

[0050] 3. According to the estimated value of the rotor position and the motor temperature, speed and torque, the above parameter information is used as input to the established neural network-based permanent magnet synchronous motor rotor position dynamic compensation module, and the motor rotor position compensation value is output.

[0051] A permanent magnet synchronous motor rotor position dynamic compensation module based on a neural network is established, including parameter setting and parameter calibration of the neural network model.

[0052] 31. Parameter setting of the neural network model, including setting the input layer, output layer and output conditions. The input layer is the motor rotor position estimation information, motor speed, motor torque, and motor temperature; the output layer is the motor rotor position compensation value. The output condition is that the difference between the rotor estimation value and the actual value is less than or equal to the rotor position compensation threshold, which is the maximum value of the motor rotor position accuracy required for precise motor control.

[0053] The calibration process of 32 parameters is as follows:

[0054] 321. Start calibration, select multiple standard motors, connect high-precision photoelectric encoders at the coaxial position of the electrode test bench, and calibrate the rotor position. The motor control uses the photoelectric encoder to transmit the decoded electronic position and speed information.

[0055] 322. During the calibration process, the three-phase current, motor temperature, stator resistance, speed and torque of the motor are collected.

[0056] 323. Calculate the estimated value of the motor rotor position.

[0057] 324. The estimated rotor position, the actual position collected by the photoelectric encoder, the motor three-phase current, the motor temperature, the stator resistance, the speed and the torque are input into the neural network permanent magnet synchronous motor rotor position dynamic compensation module to calculate the speed, torque and temperature correlation-based neural network permanent magnet synchronous motor rotor position dynamic compensation module; the optimization standard is selected as the angle difference to be kept within a reasonable threshold.

[0058] 325. The permanent magnet synchronous motor rotor position dynamic compensation module based on neural network in the previous step is verified on the test bench to ensure that the motor rotor position reliability meets the motor design requirements.

[0059] 4. Based on the rotor position compensation value and the rotor position estimation value, the rotor position value of the next carrier frequency cycle is obtained.

[0060] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for rotor position compensation without position sensor based on neural network method, characterized in that: Based on the neural network learning prediction model, the method includes the following steps: first, the three-phase current, stator resistance, inductance information, motor temperature, speed, and torque of the motor are collected and acquired; then the back electromotive force of the motor is calculated according to the collected data, and the rotor position estimation information is calculated using the back electromotive force; then the rotor position estimation information, motor rotor temperature, speed, torque and other information are input into the neural network learning prediction model, and the rotor position estimation value and compensation value are calculated to obtain the rotor position prediction value, which is used for the rotor position of the next carrier frequency cycle.

2. A method for rotor position compensation without position sensor based on neural network method according to claim 1, characterized in that: The following steps are involved: A. Collect the motor's three-phase current, motor temperature, stator resistance, speed and torque parameters to obtain more data for estimating the rotor position and confirming the motor's working condition; B. Calculate an estimated value of the motor rotor position based on the motor operation information obtained in step A; C. According to the estimated value of the rotor position and the motor temperature, speed, and torque, the above parameter information is used as input to the established neural network-based permanent magnet synchronous motor rotor position dynamic compensation module, and the motor rotor position compensation value is output; D. Based on the rotor position compensation value and the rotor position estimation value, the rotor position value of the next carrier frequency cycle is obtained.

3. The method for position sensorless rotor position compensation based on a neural network method according to claim 2, characterized in that: Step A: Parameter selection must be based on the application platform. Temperature, speed, and torque are operating condition confirmation parameters and are mandatory options. Three-phase current, stator resistance, etc. are optional options and are required by the rotor position estimation method selected by the application platform.

4. The method for position sensorless rotor position compensation based on a neural network method according to claim 2, characterized in that: Step B, adjusting the type of motor information to be collected in the first stage.

5. The method for position sensorless rotor position compensation based on a neural network method according to claim 2, characterized in that: Step C, establishing a permanent magnet synchronous motor rotor position dynamic compensation module based on a neural network, including parameter setting and parameter calibration of the neural network model.

6. The method for position sensorless rotor position compensation based on a neural network method according to claim 5, characterized in that: The parameter setting of the neural network model includes setting the input layer, output layer and output conditions.

7. The method for rotor position compensation without position sensor based on neural network method according to claim 6, characterized in that: The input layer includes motor rotor position estimation information, motor speed, motor torque, and motor temperature.

8. The method for position sensorless rotor position compensation based on a neural network method according to claim 6, characterized in that: The output layer is the motor rotor position compensation value.

9. The method for position sensorless rotor position compensation based on a neural network method according to claim 6, characterized in that: The output condition is that the difference between the rotor estimated value and the actual value is less than or equal to the rotor position compensation threshold, and the compensation threshold is the maximum value of the motor rotor position accuracy required for precise motor control.

10. The method for rotor position compensation without position sensor based on neural network method according to claim 5, characterized in that: The parameter calibration process includes the following steps: a. Coaxially connect multiple standard motors and high-precision photoelectric encoders on a motor stand, collect rotor position signals and speed signals output by the photoelectric encoder, decode the position signals and speed signals and input them into the motor control system for precise motor control; b. Collect the three-phase current, motor temperature, stator resistance, speed, and torque of the motor during the calibration process, calculate the estimated value of the rotor position and the motor operating conditions, and calculate the estimated value of the motor rotor position based on the above collected information; c. Input the estimated value of the rotor position, the actual value of the motor position calculated by the photoelectric encoder, the speed, torque, and temperature information into the dynamic compensation module of the rotor position of the permanent magnet synchronous motor based on the neural network, sum the calculated rotor position compensation value with the estimated value, and calculate the difference with the actual value solved by the photoelectric encoder to determine whether the difference meets the expected value, place a large amount of test data in the neural network learning system, and adjust the dynamic compensation module of the rotor position of the permanent magnet synchronous motor based on the neural network to meet the motor precision control requirements in the full speed range and full torque range; d. After the dynamic compensation module of the permanent magnet synchronous motor rotor position based on the neural network is adjusted, the control verification is carried out on the bench test to verify that its reliability meets the motor control requirements.

Images