CN113193789B

CN113193789B - Motor starting control parameter optimization method and device and motor starting control system

Info

Publication number: CN113193789B
Application number: CN202110589655.7A
Authority: CN
Inventors: 刘吉平; 陈筠; 王翔
Original assignee: Shenzhen Hangshun Chip Technology R&D Co Ltd
Current assignee: Shenzhen Hangshun Chip Technology R&D Co Ltd
Priority date: 2021-05-27
Filing date: 2021-05-27
Publication date: 2022-07-26
Anticipated expiration: 2041-05-27
Also published as: CN113193789A

Abstract

The application discloses a method and a device for optimizing motor starting control parameters, a storage medium, motor configuration equipment and a motor starting control system. The method comprises the following steps: randomly generating motor starting control parameters according to a preset value range of the motor starting control parameters in the motor starting model, and establishing a training set and a test set based on the randomly generated motor starting control parameters; establishing a deep learning network model, training the deep learning network model by using a training set, and verifying the trained model by using a testing set to obtain a target deep learning network model; and inputting the motor starting control parameters in the target motor sample set into the target deep learning network model, and acquiring the optimal motor starting control parameters of the target motor according to the motor starting predicted value output by the target deep learning network model. The motor control method and the motor control system can be suitable for various motors of different types and use scenes, and can enable performance indexes of all motors to be optimally controlled.

Description

Motor starting control parameter optimization method and device and motor starting control system

Technical Field

The application relates to the technical field of motor control, in particular to a method and a device for optimizing motor starting control parameters, a storage medium, motor configuration equipment and a motor starting control system.

Background

With the rapid development of motor technology, the dc brushless motor has been widely used in automobiles, tools, home appliances, industrial control, automation, aerospace industry, and the like. The operation of a dc brushless motor requires a dedicated drive algorithm to control and protect the motor. In the related art, when the motor is controlled, it is difficult to find a perfect motor start control parameter, so that the motor start control parameter is suitable for various motors and use scenes, and therefore, the performance index of each motor cannot be optimally controlled.

Disclosure of Invention

In view of this, the present application provides a method, an apparatus, a storage medium, a motor configuration device, and a motor start control system for optimizing a motor start control parameter, so as to solve the problem that the motor start control parameter in the existing scheme is not suitable for various types of motors and use scenarios, and cannot enable performance indexes of each type of motor to be optimally controlled.

In a first aspect, an embodiment of the present application provides a method for optimizing a motor start control parameter, including:

randomly generating motor starting control parameters according to a preset value range of the motor starting control parameters in the motor starting model, and establishing a training set and a testing set based on the randomly generated motor starting control parameters;

establishing a deep learning network model, training the deep learning network model by using the training set, and verifying the trained model by using the test set to obtain a target deep learning network model;

and inputting the motor starting control parameters in the sample set of the target motor into the target deep learning network model, and acquiring the optimal motor starting control parameters of the target motor according to the motor starting predicted value output by the target deep learning network model.

Optionally, the motor start control parameters include a motor start input parameter and a motor start target value, the motor start control parameters are randomly generated according to a preset value range of the motor start control parameters in the motor start model, and a training set and a test set are established based on the randomly generated motor start control parameters, including:

the motor starting input parameters are randomly generated according to the preset value range of the motor starting input parameters in the motor starting model, and online observation data are generated through online monitoring;

determining the motor starting target value according to the online observation data;

and establishing the training set and the testing set according to the motor starting input parameters and the motor starting target values.

Optionally, the establishing a training set and a test set according to the motor start input parameter and the motor start target value includes:

if the distribution of the motor starting input parameters and the motor starting target values meets preset conditions, normalization processing is carried out on the motor starting input parameters and the motor starting target values;

and establishing the training set and the testing set according to the motor starting input parameters and the motor starting target values after normalization processing.

Optionally, the motor start input parameters include at least:

duration of the dc excitation period, initial value of the reference current during the dc excitation period, final value of the reference current during the dc excitation period, maximum duration of the forced commutation period, rotor speed at the end of the preset total time, and PID (proportional-integral-derivative) parameters of the rotor speed to current.

Optionally, the motor start target value at least includes any one of a motor start power target value and a motor start time target value.

Optionally, when the motor start target value includes the motor start power target value, the online observation data includes real-time monitoring of total current, time from rest to start of movement of the rotor, and time from start of movement of the rotor to the time at which the speed thereof reaches the target speed, the determining the motor start target value according to the online observation data includes:

and determining the starting power target value of the motor according to the input voltage of the motor, the real-time monitoring total current and the sum of the time from the standstill to the start of the movement of the rotor and the time from the start of the movement of the rotor to the time that the speed of the rotor reaches the target speed.

Optionally, when the motor start target value includes the motor start time target value, the online observation data includes a time from a standstill of the rotor to a start of movement and a time from the start of movement of the rotor to a time at which a speed thereof reaches a target speed, and the determining the motor start target value according to the online observation data includes:

and determining the target value of the starting time of the motor according to the time from the standstill of the rotor to the start of movement and the time from the start of movement of the rotor to the time when the speed of the rotor reaches the target speed.

Optionally, when the motor start target value includes the motor start power target value and the motor start time target value, the online observation data includes real-time monitoring of a total current, a time from a standstill of the rotor to a start of movement of the rotor, and a time from the start of movement of the rotor to a time at which a speed thereof reaches a target speed, the determining the motor start target value according to the online observation data includes:

determining a motor starting power target value according to the motor input voltage, the real-time monitoring total current and the sum of the time from the standstill of the rotor to the start of movement of the rotor and the time from the start of movement of the rotor to the time when the speed of the rotor reaches the target speed; and

Optionally, the inputting the motor start control parameters in the sample set of the target motor into the target deep learning network model, and obtaining the optimal motor start control parameters of the target motor according to the motor start predicted value output by the target deep learning network model includes:

preprocessing the motor starting control parameters in the sample set of the target motor according to a preset searching method;

inputting motor starting control parameters in a sample set of the preprocessed target motor into the target deep learning network model;

obtaining an optimal motor starting predicted value from the motor starting predicted values output by the target deep learning network model;

and acquiring a motor starting control parameter corresponding to the optimal motor starting predicted value, and taking the motor starting control parameter corresponding to the optimal motor starting predicted value as the optimal motor starting control parameter.

Optionally, before randomly generating a motor start control parameter according to a preset value range of the motor start control parameter in the motor start model and establishing a training set and a test set based on the randomly generated motor start control parameter, the method further includes:

and selecting the motor starting model.

In a second aspect, an embodiment of the present application provides a motor start control parameter optimization device, including:

the device comprises a first establishing module, a second establishing module and a third establishing module, wherein the first establishing module is used for randomly generating motor starting control parameters according to a preset value range of the motor starting control parameters in a motor starting model and establishing a training set and a test set based on the randomly generated motor starting control parameters;

the second establishing module is used for establishing a deep learning network model, training the deep learning network model by using the training set and verifying the trained model by using the testing set so as to obtain a target deep learning network model;

and the acquisition module is used for inputting the motor starting control parameters in the sample set of the target motor into the target deep learning network model and acquiring the optimal motor starting control parameters of the target motor according to the motor starting predicted value output by the target deep learning network model.

In a third aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed on a computer, the computer is caused to execute the flow in the motor starting control parameter optimization method provided in the present application.

In a fourth aspect, an embodiment of the present application provides a motor configuration device, which includes a memory and a processor, where the processor executes a flow in the motor start control parameter optimization method provided in the embodiment of the present application by calling a computer program stored in the memory.

In a fifth aspect, an embodiment of the present application further provides a motor start control system, including main control end and the motor drive board that are connected, the main control end is used for generating a test set, establishing a deep learning network model and generating motor start control parameters, be equipped with motor drive chip on the motor drive board, motor drive chip is right the motor start control parameters that the main control end sent carry out the start test, and the control motor rotates, collects state parameters in the motor start process and sends to the main control end carries out the analysis, so that the main control end determines the optimal motor start control parameters.

In the method and apparatus for optimizing motor start control parameters, the storage medium, the motor configuration device, and the motor start control system according to the embodiments of the present application, first, motor start control parameters may be randomly generated according to a preset value range of the motor start control parameters in a selected motor start model, and a training set and a test set may be established based on the randomly generated motor start control parameters. And then, establishing a deep learning network model, training the deep learning network model by using a training set, and verifying the trained model by using a testing set, so that the target deep learning network model can be obtained. And then, inputting the motor starting control parameters in the sample set of the target motor into the target deep learning network model, and acquiring the optimal motor starting control parameters of the target motor according to the motor starting predicted value output by the target deep learning network model. The motor starting control parameters are constructed by adopting a deep learning method, and corresponding motor starting control parameters can be constructed aiming at different motors in a self-adaptive manner. Therefore, the motor control method and the motor control device can be suitable for various motors of different types and use scenes, and the performance indexes of all the motors can be optimally controlled.

Drawings

The technical solutions and advantages of the present application will be apparent from the following detailed description of specific embodiments of the present application with reference to the accompanying drawings.

Fig. 1 is a schematic flowchart of a method for optimizing a motor start control parameter according to an embodiment of the present application;

fig. 2 is a schematic view of a scenario of setting a motor start control parameter provided in an embodiment of the present application;

FIG. 3 is a schematic view of a scene of online observation of a motor starting process provided in an embodiment of the present application;

FIG. 4 is a schematic diagram of a scene of a deep neural network model provided by an embodiment of the present application;

fig. 5 is a schematic structural diagram of a motor start control parameter optimization device provided in an embodiment of the present application;

fig. 6 is a schematic structural diagram of a motor configuration apparatus provided in an embodiment of the present application;

fig. 7 is a schematic structural diagram of a motor start control system according to an embodiment of the present application.

Detailed Description

Reference is made to the drawings, wherein like reference numerals refer to like elements, which are illustrated in the various figures, and which are implemented in a suitable computing environment. The following description is based on illustrated embodiments of the application and should not be taken as limiting the application with respect to other embodiments that are not detailed herein.

It is understood that the execution subject of the embodiment of the present application may be a motor configuration apparatus. In practical scenarios, the embodiment of the present application is not limited to the specific representation form of the motor configuration apparatus, and includes, but is not limited to: the mobile phone, the tablet computer, the Personal Computer (PC), the cloud computer, and the like have corresponding functions.

Referring to fig. 1, fig. 1 is a schematic flow chart of a method for optimizing a motor start control parameter according to an embodiment of the present disclosure. The process of the motor starting control parameter optimization method can comprise the following steps:

101. randomly generating motor starting control parameters according to a preset value range of the motor starting control parameters in the motor starting model, and establishing a training set and a testing set based on the randomly generated motor starting control parameters.

With the rapid development of motor technology, the brushless dc motor has been widely used in automobiles, tools, home appliances, industrial control, automation, aerospace industry, etc. A control system of a brushless direct current motor relates to the subjects of power electronic technology, electromechanics, automatic control, material science and the like. The operation of a dc brushless motor requires a dedicated drive algorithm to control and protect the motor. With the development of motor technology towards integration, intellectualization, high efficiency and energy conservation, the market scale of the motor technology is further expanded.

The following problems generally exist when a dc brushless motor is started: 1. different motors have own rotational inertia, no comparability exists between the motors, and individuation of different application scenes is strong. 2. Improper motor start control parameters cause many problems because the start impact current can reach 4-7 times of rated current, the motor current heats seriously, and continuous start and stop can cause heat accumulation and motor burnout. 3. The starting conditions of different motors are different, some motors are provided with Hall sensors, position sensors such as encoders, etc., some motors are not provided with position sensors, and other motors are provided with resistance sensors (such as a sensor for measuring water pressure in a water pump, a sensor for measuring weight in a washing machine, etc.).

For the reasons, for each type of motor, the motor starting control parameters need to be calculated according to the parameters of the motor and the parameters of the control circuit, the motor starting control parameters are adjusted according to the starting state of the motor, including starting time, starting current, starting jitter and the like, the test is performed again, and the steps are repeated and basically adopt an exhaustion method. In addition, even with an expert with industry experience, the task of manually optimizing the model parameters takes a lot of time. And the estimation of the motor starting state is possibly insufficient, so that the motor starting control parameters cannot be adapted to all environments.

As can be seen from the above, in the related art, when controlling the motor, under given constraint conditions, it is difficult to find a perfect motor start control parameter, so that the motor start control parameter is suitable for various types of motors and use scenarios, and the performance index of the motor control system is optimally controlled.

In the embodiment of the application, for motors in different application scenes (use scenes), the corresponding motor starting models are different, so that when motor starting control parameters are optimized, the corresponding motor starting models are selected according to the current application scene.

It can be understood that, if the corresponding motor starting model is selected in advance according to the current application scenario before the step 101 is executed, the embodiment of the present application is executed directly from the step 101. If the corresponding motor starting model is not selected before the step 101 is executed, the embodiment of the present application first selects the corresponding motor starting model according to the current application scenario, and then executes the step 101.

In the embodiment of the present application, a motor start parameter set may be defined according to a motor start control parameter in a motor start model. And then, according to the significance of the motor starting control parameters in the motor starting parameter set, giving the value range of the motor starting control parameters. Then, the motor start control parameters can be randomly generated according to the preset value range of the motor start control parameters in the motor start model, that is, the motor start control parameters are randomly generated within the defined preset value range of the motor start control parameters, and a training set and a test set can be established based on the randomly generated motor start control parameters. For example, 100 sample sets may be established, wherein 80 sample sets are used as a training set, and the remaining 20 sample sets are used as a test set, where all samples in the training set and the test set are motor start control parameters, but values of the motor start control parameters in each of the training set and the test set are different.

In one embodiment, the motor start control parameters may include a motor start input parameter and a motor start target value, and the randomly generating 101 motor start control parameters according to a preset value range of the motor start control parameters in the motor start model, and establishing a training set and a testing set based on the randomly generated motor start control parameters may include:

randomly generating motor starting input parameters according to a preset value range of the motor starting input parameters in the motor starting model, and generating online observation data through online monitoring;

determining a motor starting target value according to the online observation data;

and establishing a training set and a test set according to the motor starting input parameters and the motor starting target values.

For example, the motor start input parameters may be randomly generated according to a preset value range of the motor start input parameters in the motor start model, that is, the motor start input parameters are randomly generated within a defined preset value range of the motor start control parameters, and online observation data may be generated through online monitoring. The motor starting target value can be determined according to the online observation data.

For example, in one embodiment, the motor start input parameters may include at least: the duration of the dc excitation period, the initial value of the reference current during dc excitation, the final value of the reference current during dc excitation, the maximum duration of the forced commutation period, the speed of the rotor (i.e. the rotor of the electrical machine) at the end of the preset total time, and the PID parameter of the rotor speed converted into current.

It should be noted that, when the pressure sensor is installed in the motor, the motor start input parameter may further include an initial pressure value of the pressure sensor, and the like.

For example, in one embodiment, the motor start target value may include at least any one of a motor start power target value and a motor start time target value. That is, the motor start target value may include a motor start power target value, or the motor start target value may include a motor start time target value, or the motor start target value may include a motor start power target value and a motor start time target value.

For example, in an embodiment, when the motor start target value includes a motor start power target value, the description is focused on the thermal energy consumption at the motor start, and it is desirable to reduce the thermal energy consumption at the motor start to achieve the purpose of optimizing the motor start, in this case, the online observation data may include real-time monitoring of the total current, the time from the rotor being stationary to the rotor starting to move, and the time from the rotor starting to the speed of the rotor reaching the target speed, and the determining the motor start target value according to the online observation data may include:

For example, the input voltage of the motor is a constant voltage Vbus, the product of the constant voltage Vbus and the real-time monitored total current I is integrated, and the length of the integration is t0+ t1, where t0 is the time from the standstill of the rotor to the start of movement, and t1 is the time from the start of movement of the rotor to the time when the speed of the rotor reaches the target speed, so that a motor starting power target value Ps can be obtained, and a calculation formula thereof is as follows:

for example, in an embodiment, when the motor start target value includes a motor start time target value, indicating that the start time of the motor is relatively concerned, it is desirable to reduce the start time of the motor for the purpose of optimizing the motor start, in which case, the online observation data may include a time from the standstill of the rotor to the start of the rotor and a time from the start of the rotor to the time at which the speed of the rotor reaches the target speed, and the determining the motor start target value according to the online observation data may include:

For example, the sum of the time T0 from the standstill of the rotor to the start of the movement and the time T1 from the start of the movement of the rotor to the time at which the speed thereof reaches the target speed is the motor start time target value T.

For example, in an embodiment, when the motor start target value includes a motor start power target value and a motor start time target value, it is noted that the heat energy consumption and the start time of the motor are concerned, and it is desirable to reduce the heat energy consumption and the start time of the motor when the motor is started to achieve the purpose of optimizing the motor start, in this case, the online observation data may include real-time monitoring of the total current, the time from the standstill to the start of the rotor, and the time from the start of the rotor to the time until the speed of the rotor reaches the target speed, and the determining the motor start target value according to the online observation data may include:

For example, the input voltage of the motor is a constant voltage Vbus, the product of the constant voltage Vbus and the real-time monitoring total current I is integrated, and the length of the integration is t0+ t1, where t0 is the time from the standstill of the rotor to the start of movement, and t1 is the time from the start of movement of the rotor to the time when the speed of the rotor reaches the target speed, so that the motor starting power target value Ps can be obtained, and the calculation formula is as follows:

the time t0 from the standstill of the rotor to the start of the movement is added to the time t1 from the start of the movement of the rotor to the time when the speed of the rotor reaches the target speed, and the sum is the target motor starting time value Ts, i.e., Ts is t0+ t 1.

In one embodiment, the creating a training set and a test set according to the motor start input parameters and the motor start target values may include:

if the distribution of the motor starting input parameters and the motor starting target values meets preset conditions, performing normalization processing on the motor starting input parameters and the motor starting target values;

For example, the preset condition is that the distribution is not uniform, and if the distribution of the motor start input parameter and the motor start target value satisfies the preset condition, the distribution of the motor start input parameter and the motor start target value is not uniform, in other words, the distribution of the motor start input parameter and the motor start target value is relatively dispersed and is not concentrated in one area. At this time, data standardization is required, for example, normalization processing can be performed on the motor starting input parameter and the motor starting target value, so that the motor starting input parameter and the motor starting target value after normalization processing are both located between 0 and 1 or between-1 and 1, and thus, the motor starting input parameter and the motor starting target value can be uniformly distributed. If the motor starting input parameters and the motor starting target values are uniformly distributed, normalization processing is not needed.

After the motor start input parameters and the motor start target values are normalized, a training set and a test set can be established according to the normalized motor start input parameters and the normalized motor start target values. For example, 100 sample sets may be established, wherein 80 sample sets are used as training sets, and the remaining 20 sample sets are used as test sets, where the samples in the training sets and the test sets are motor start input parameters and motor start target values after normalization processing, but values of the motor start input parameters and the motor start target values after normalization processing in each of the training sets and the test sets are different.

102. And establishing a deep learning network model, training the deep learning network model by using a training set, and verifying the trained model by using a testing set to obtain the target deep learning network model.

In the embodiment of the application, a deep learning network model is established, and the deep learning network model is trained by using a training set. For example, the deep learning network model is trained using 80 training sets, and the motor start input parameters and the motor start target values in the 1 st training set are input into the deep learning network model, and the deep learning network model outputs the motor start prediction values. The square of the difference between the predicted motor start value and the target motor start value can be calculated by using a Mean Squared Error (MSE) loss function. And if the square of the difference between the motor starting predicted value and the motor starting target value is greater than a preset mean square error threshold value, adjusting the structure of the deep learning network model according to the square of the difference between the motor starting predicted value and the motor starting target value. And then, inputting the motor starting input parameters and the motor starting target values in the 2 nd training set into the adjusted deep learning network model, then solving the square of the difference between the motor starting predicted value and the motor starting target value output by the adjusted deep learning network model by adopting a mean square error loss function, and so on until the 80 th training set is used for training. And training by the 80 th training set to obtain a trained model.

After the trained model is obtained, the performance of the trained model needs to be verified by using a test set. For example, the trained model is verified using 20 test sets. In the process of verifying the trained model, the Error between the motor starting predicted value output by the model trained in each verification process and the motor starting target value is obtained, so that 20 errors are obtained, and then the Mean Absolute Error (MAE) of the 20 errors is obtained, namely the Mean Absolute Error (MAE) of the 20 errors is obtained, so as to evaluate the performance of the trained model.

And if the average absolute error is larger than the preset error threshold, retraining the trained model until the average absolute error is smaller than or equal to the preset error threshold, and obtaining the final target deep learning network model.

103. And inputting the motor starting control parameters in the sample set of the target motor into the target deep learning network model, and acquiring the optimal motor starting control parameters of the target motor according to the motor starting predicted value output by the target deep learning network model.

It can be understood that, in the embodiment of the application, according to actual requirements, the motor start control parameters are input into the target deep learning network model, and the optimal motor start control parameters of the target motor are selected by adopting a preset search method. It should be noted that the target motor refers to a motor for which an optimal motor start control parameter needs to be predicted at present.

For example, in an embodiment, the inputting the motor start control parameters in the sample set of the target motor into the target deep learning network model and obtaining the optimal motor start control parameters of the target motor according to the predicted motor start values output by the target deep learning network model may include:

For example, in one embodiment, when the preset search method is a grid search method, the motor start control parameter needs to be subjected to grid division, that is, the duration of the dc excitation period, the initial value of the reference current during the dc excitation period, the final value of the reference current during the dc excitation period, the maximum duration of the forced commutation period, the rotor speed at the end of the preset total time, the PID parameter of the rotor speed converted into the current, and the motor start target value are subjected to grid division within respective value ranges by a preset division amount, so as to generate more sample sets, and improve the calculation accuracy of the optimal motor start control parameter. For example, if the value range of the duration of the dc excitation period is [0,10], and the preset step is 0.5, the following parameter sequence of the duration of the dc excitation period can be obtained: (0, 0.5, 1, 1.5 … … 9.5.5, 10), namely, the value range of the duration of the direct current excitation period is divided into 20 grids, and the step size of each grid is the same.

In analogy, the initial value of the reference current during the current excitation, the final value of the reference current during the direct current excitation, the maximum duration of the forced commutation period, the rotor speed at the end of the preset total time, the PID parameter for converting the rotor speed into the current, and the motor start target value may also be divided into 20 grids within their respective value ranges.

Inputting the duration of a direct-current excitation period, an initial value of a reference current of the direct-current excitation period, a final value of the reference current of the direct-current excitation period, a maximum duration of a forced commutation period, a rotor speed at the end of a preset total time, a PID (proportion integration differentiation) parameter of which the rotor speed is converted into a current and a motor starting target value into a target deep learning network model, wherein the duration, the initial value, the final value, the maximum duration, the rotor speed and the PID parameter are obtained by grid division, and the motor starting control parameter is corresponding to the minimum motor starting power predicted value or the minimum motor starting time predicted value.

It will be appreciated that in the grid search, the finer the grid division, the closer to the optimal motor start control parameter, but the more computing resources are consumed. For example, dividing the initial value of the reference current during the current excitation, the final value of the reference current during the dc excitation, the maximum duration of the forced commutation period, the rotor speed at the end of the preset total time, the PID parameter of the rotor speed converted into the current, and the motor start target value into 40 grids within the respective value ranges, the higher the accuracy of the calculated optimal motor start control parameter is, but more calculation resources are consumed.

For example, in another embodiment, when the preset search method is a random search method, a value range is preset for the motor start control parameter, the motor start control parameter is randomly generated within the preset value range, and the sample set is randomly generated. For example, a sample set is randomly generated by randomly generating the duration of the direct current excitation period, the initial value of the reference current during the direct current excitation period, the final value of the reference current during the forced commutation period, the rotor speed at the end of the preset total time, the PID parameter of the rotor speed converted into the current and the preset value range of the motor starting target value within the preset value range.

And inputting the motor starting control parameters in the randomly generated sample set into the target deep learning network model, finding an optimal value from the motor starting predicted values output by the target deep learning network model, for example, finding a minimum motor starting power predicted value or a minimum motor starting time predicted value, and taking the motor starting control parameters corresponding to the minimum motor starting power predicted value or the minimum motor starting time predicted value as the optimal motor starting control parameters.

For example, in other embodiments, when the preset search method is a grid search + bayesian optimization search method, the search process may be divided into two stages, the first stage is a grid search, and when the grid search is performed, the grid step size may be larger, that is, the number of the grid divisions may be smaller, for example, the motor start control parameter may be divided into 10 grids, that is, the duration of the dc excitation period, the initial value of the reference current during the dc excitation period, the final value of the reference current during the dc excitation period, the maximum duration of the forced commutation period, the rotor speed at the end of the preset total time, the PID parameter of the rotor speed converted into the current, and the motor start target value are respectively divided into 10 grids. And obtaining an optimal motor starting control parameter after grid search.

In order to improve the searching precision, a Bayesian optimization searching method is adopted near the optimal motor starting control parameter searched by the grid, namely, the optimal motor starting control parameter with higher precision is detected in the area near the optimal motor starting control parameter searched by the grid.

And taking the area near the optimal motor starting control parameter searched by the grid as a finer searching range, namely reducing the value range of the motor starting control parameter, reducing the value range to the area near the optimal motor starting control parameter searched by the grid, and carrying out grid division on the area near the optimal motor starting control parameter by using smaller step length so as to divide the area into more grids. For example, the area near the optimal motor start control parameter searched by the grid is divided into 30 grids, and the optimal motor start control parameter is detected in a finer range, so that the accuracy of the calculated optimal motor start control parameter is improved.

It should be noted that the bayesian optimization search method finds a value that minimizes the objective function by establishing a substitute function (probability model) based on the past evaluation result of the objective function. The Bayesian optimization search method is different from the random search method or the grid search method in that the Bayesian optimization search method refers to the previous evaluation result when trying the next set of hyper-parameters, so that much useless work can be saved.

It can be understood that, in the embodiment of the present application, first, the motor start control parameter may be randomly generated according to a preset value range of the motor start control parameter in the selected motor start model, and a training set and a test set may be established based on the randomly generated motor start control parameter. And then, establishing a deep learning network model, training the deep learning network model by using a training set, and verifying the trained model by using a testing set, so that the target deep learning network model can be obtained. And then, inputting the motor starting control parameters in the sample set of the target motor into the target deep learning network model, and acquiring the optimal motor starting control parameters of the target motor according to the motor starting predicted value output by the target deep learning network model. The motor starting control parameters are constructed by adopting a deep learning method, and corresponding motor starting control parameters can be constructed in a self-adaptive mode aiming at different motors. Therefore, the motor control method and the motor control device can be suitable for various motors of different types and use scenes, and can enable the performance indexes of all the motors to be optimally controlled.

The motor starting control parameter optimization method is configured in the motor configuration tool, so that the motor configuration tool automatically adapts to the motor starting control parameters, debugging cost can be effectively reduced, complexity of debugging motor starting can be reduced, and the motor starting control parameters are more suitable for application scenes.

Referring to fig. 2, fig. 2 is a schematic view of a scenario of setting a motor start control parameter according to an embodiment of the present application. In a scene of a sensorless Permanent Magnet Synchronous Motor (PMSM), a torque starting model is adopted as a selected Motor starting model.

The starting of a sensorless dc brushless motor generally employs the following three steps:

1) during the direct-current excitation: during the dc excitation period, the rotor is rotated to a fixed position in order to force a current to flow into the coil. During this dc excitation, most of the electrical energy generates a large amount of heat.

The input parameters during the dc excitation are shown in fig. 2, the duration is 0-1000ms, a is a current variation curve, b is a rotor speed variation curve, and the corresponding parameters are as follows:

duration of the dc excitation period: 1000 ms;

initial value of reference current during dc excitation: 0A;

final value of reference current during dc excitation: 1.2A.

2) During forced commutation: the rotor in a stopped state is gradually applied with a rotating magnetic field to start rotating. As shown in fig. 2, the interval with the duration of 1000ms to 4000ms is a forced commutation period, and a forced commutation energization signal may be applied to the motor driving chip at a certain frequency to rotate the motor.

The parameters during forced commutation are as follows:

maximum duration during forced commutation: 3000 ms;

rotor speed at the end of the total time allowed: 3000 rpm;

the speed translates into a PID parameter of the current.

It should be noted that the current and the torque have a corresponding relationship, and the magnitude of the torque can be calculated after the current is acquired.

In some embodiments, the parameters during forced commutation may also include a maximum current that allows the generation of a three-phase current system that initiates the rotating stator flux: 1.67A, the maximum current of the three-phase current system which allows the generation of the starting rotating stator magnetic flux has the function of limiting the current of the three-phase current system which allows the generation of the starting rotating stator magnetic flux not to exceed the maximum current of the three-phase current system which allows the generation of the starting rotating stator magnetic flux, and the motor is protected.

3) Detection of motor induced voltage: during forced commutation, the motor starts to rotate and the phase coils generate induced voltages. When the induced voltage is input to the position signal input terminal, the position of the rotor is known. And when the position of the motor is judged to be credible, closed-loop control is adopted.

Referring to fig. 3, fig. 3 is a scene schematic diagram of online observation of a motor starting process according to an embodiment of the present application. And c is a current change curve in the motor starting process observed on line.

In the above motor starting model, a motor starting input parameter type and an online observation data type are defined, and raw data is generated. The motor starting input parameters comprise duration T0 of a direct current excitation period, an initial value C _ init of a reference current of the direct current excitation period, a final value C _ last of the reference current of the direct current excitation period, a maximum duration T1 of a forced commutation period, a rotor Speed _ last at the end of an allowed total time and a PID parameter PID _ para of Speed conversion to current, and online observation data comprise real-time monitoring of the total current I, time T0 from the standstill of a rotor to the start of movement of the rotor and time T1 from the start of the rotor to the time T1 of the Speed of the rotor to reach a target Speed. The motor starting input parameters and the online observation data form original data.

And when the motor starting target value is the motor starting power target value, processing the original data and calculating the motor starting power target value Ps. The calculation formula is as follows:

vbus is the input voltage of the motor, I is the real-time monitored total current, t0 is the time from standstill to the start of the movement of the rotor, t1 is the time from the start of the movement of the rotor to its speed reaching the target speed.

Then, a sample set may be established: [ T0, C _ init, C _ last, T1, Speed _ last, PID _ para, Ps ]. For example, 100 sample sets are established, wherein 80 sample sets are used as training sets, and the remaining 20 sample sets are used as testing sets.

Thereafter, a deep learning network model can be constructed. Referring to fig. 4, fig. 4 is a schematic view of a deep neural network model according to an embodiment of the present disclosure. Due to the small number of samples, a very small network can be used, the deep neural network model includes 1 input layer, 2 hidden layers and 1 output layer, with a typical setup of a scalar regression model.

When the deep neural network model is trained, the square of the difference between the motor starting predicted value and the motor starting target value is calculated by adopting a mean square error loss function. And verifying the trained model by adopting a test set, and solving an average absolute error in the test process, namely solving an average value of errors between the motor starting predicted value and the motor starting target value obtained by multiple tests to evaluate the performance of the trained model.

Finally, acquiring an optimal motor starting predicted value from the motor starting predicted values output by the target deep learning network model by adopting a preset searching method; and then obtaining a motor starting control parameter corresponding to the optimal motor starting predicted value, and taking the motor starting control parameter corresponding to the optimal motor starting predicted value as the optimal motor starting control parameter.

For example, when performing grid search, the motor start control parameters are divided into 20 grids in the value range, and if the duration T0 of the dc excitation period is divided into 20 grids and the sizes of the corresponding 20 motor start power predicted values are checked, the duration T0 of the dc excitation period may be used as an index, a grid search method is adopted to generate 20 motor start power predicted values, and the motor start control parameter corresponding to the minimum motor start power predicted value is selected as the optimal motor start control parameter.

It can be understood that, in a scenario where the motor is equipped with a pressure sensor, the motor start input parameter may further include an initial pressure value Pint of the pressure sensor, and correspondingly, the sample set may also include the initial pressure value Pint of the pressure sensor. By the motor starting control parameter optimization method in the embodiment of the application, the optimal motor starting control parameter corresponding to the scene can be generated.

Similarly, when the motor is equipped with a hall sensor or encoder, then the motor start input parameters may also include the position of the rotor, at which time the cost may increase. When the motor is equipped with a weight sensor, then the motor start input parameter may also include weight. When the motor is equipped with a water pressure sensor, then the motor start input parameters may also include an initial value of water pressure.

Referring to fig. 5, fig. 5 is a schematic structural diagram of a motor start control parameter optimization device according to an embodiment of the present disclosure. The motor start-up control parameter optimizing device 200 may include: a first establishing module 201, a second establishing module 202 and an obtaining module 203.

The first establishing module 201 is configured to randomly generate a motor start control parameter according to a preset value range of the motor start control parameter in a motor start model, and establish a training set and a test set based on the randomly generated motor start control parameter;

a second establishing module 202, configured to establish a deep learning network model, train the deep learning network model using the training set, and verify the trained model using the test set to obtain a target deep learning network model;

the obtaining module 203 is configured to input the motor start control parameters in the sample set of the target motor into the target deep learning network model, and obtain the optimal motor start control parameters of the target motor according to the motor start predicted value output by the target deep learning network model.

In one embodiment, the motor start control parameters include a motor start input parameter and a motor start target value, and the first establishing module 201 may be configured to:

In one embodiment, the first establishing module 201 may be configured to:

and establishing the training set and the test set according to the motor starting input parameters and the motor starting target values after normalization processing.

In one embodiment, the motor start input parameters include at least:

the duration of the direct current excitation period, the initial value of the reference current during the direct current excitation period, the final value of the reference current during the direct current excitation period, the maximum duration of the forced commutation period, the rotor speed at the end of the preset total time, and the PID parameter of the rotor speed converted into the current.

In one embodiment, the motor start target value includes at least any one of a motor start power target value and a motor start time target value.

In one embodiment, when the motor start target value comprises the motor start power target value, the online observation data comprises real-time monitoring of a total current, a time from a standstill of the rotor to a start of the rotor, and a time from the start of the rotor to a speed of the rotor reaching a target speed, the first establishing module 201 may be configured to:

and determining the starting power target value of the motor according to the input voltage of the motor, the real-time monitoring total current and the sum of the time from the standstill of the rotor to the start of the rotor and the time from the start of the rotor to the time when the speed of the rotor reaches the target speed.

In one embodiment, when the motor start target value comprises the motor start time target value, the online observation data comprises a time from rest to start moving of the rotor and a time from start moving of the rotor to a time when the speed of the rotor reaches a target speed, and the first establishing module 201 may be configured to:

and determining the motor starting time target value according to the time from the standstill of the rotor to the start of the movement of the rotor and the time from the start of the movement of the rotor to the time when the speed of the rotor reaches the target speed.

In one embodiment, when the motor start target value includes the motor start power target value and the motor start time target value, the online observation data includes real-time monitoring of a total current, a time from a standstill of the rotor to a start of movement of the rotor, and a time from the start of movement of the rotor to a speed thereof reaching a target speed, the first establishing module 201 may be configured to:

In one embodiment, the obtaining module 203 may be configured to:

In one embodiment, the first establishing module 201 may be configured to:

and selecting the motor starting model.

The embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed on a computer, the computer is caused to execute the flow in the motor starting control parameter optimization method provided by the embodiment.

The embodiment of the present application further provides a motor configuration device, which includes a memory and a processor, where the processor is configured to execute the flow in the motor start control parameter optimization method provided in this embodiment by calling the computer program stored in the memory.

For example, the motor configuration device may be a terminal device having a corresponding function, such as a mobile phone, a tablet computer, a personal computer, a cloud computer, and the like. Referring to fig. 6, fig. 6 is a schematic structural diagram of a motor configuration apparatus according to an embodiment of the present application.

The motor configuration device 300 may include components such as a memory 301, a processor 302, and the like. Those skilled in the art will appreciate that the motor configuration device configuration shown in fig. 6 does not constitute a limitation of the motor configuration device and may include more or fewer components than those shown, or some components may be combined, or a different arrangement of components.

The memory 301 may be used to store applications and data. The memory 301 stores applications containing executable code. The application programs may constitute various functional modules. The processor 302 executes various functional applications and data processing by running the application programs stored in the memory 301.

The processor 302 is a control center of the motor configuration device, connects various parts of the motor configuration device by using various interfaces and lines, and executes various functions and processes data of the motor configuration device by running or executing the application program stored in the memory 301 and calling the data stored in the memory 301, thereby performing overall monitoring on the motor configuration device.

In this embodiment, the processor 302 in the motor configuration device loads the executable code corresponding to the processes of one or more application programs into the memory 301 according to the following instructions, and the processor 302 runs the application programs stored in the memory 301, so as to execute:

In an embodiment, the motor start control parameters include a motor start input parameter and a motor start target value, and when executing the randomly generating motor start control parameters according to a preset value range of the motor start control parameters in the motor start model, and building a training set and a test set based on the randomly generated motor start control parameters, the processor 302 may further execute: the motor starting input parameters are randomly generated according to the preset value range of the motor starting input parameters in the motor starting model, and online observation data are generated through online monitoring; determining the motor starting target value according to the online observation data; and establishing the training set and the testing set according to the motor starting input parameters and the motor starting target values.

In one embodiment, when the processor 302 executes the establishing of the training set and the test set according to the motor start input parameter and the motor start target value, it may further execute: if the distribution of the motor starting input parameters and the motor starting target values meets preset conditions, performing normalization processing on the motor starting input parameters and the motor starting target values; and establishing the training set and the test set according to the motor starting input parameters and the motor starting target values after normalization processing.

In one embodiment, the motor start input parameters include at least: the duration of the direct current excitation period, the initial value of the reference current during the direct current excitation period, the final value of the reference current during the direct current excitation period, the maximum duration of the forced commutation period, the rotor speed at the end of the preset total time, and the PID parameter of the rotor speed converted into the current.

In one embodiment, when the motor start target value includes the motor start power target value, the online observation data includes real-time monitoring of total current, time from standstill to start of movement of the rotor, and time from start of movement of the rotor to the time at which the speed of the rotor reaches the target speed, and the processor 302, in performing the determining of the motor start target value according to the online observation data, may further perform: and determining the starting power target value of the motor according to the input voltage of the motor, the real-time monitoring total current and the sum of the time from the standstill of the rotor to the start of the rotor and the time from the start of the rotor to the time when the speed of the rotor reaches the target speed.

In one embodiment, when the motor start target value includes the motor start time target value, the online observation data includes a time from a standstill to a start of movement of the rotor and a time from the start of movement of the rotor to a time at which a speed thereof reaches a target speed, and the processor 302, when performing the determining the motor start target value according to the online observation data, may further perform: and determining the target value of the starting time of the motor according to the time from the standstill of the rotor to the start of movement and the time from the start of movement of the rotor to the time when the speed of the rotor reaches the target speed.

In one embodiment, when the motor start target value includes the motor start power target value and the motor start time target value, the online observation data includes real-time monitoring of total current, time from the rotor being stationary to starting to move, and time from the rotor starting to move until the speed of the rotor reaches the target speed, and the processor 302, when executing the determining the motor start target value according to the online observation data, may further execute: determining a starting power target value of the motor according to the input voltage of the motor, the real-time monitoring total current and the sum of the time from the standstill to the start of the movement of the rotor and the time from the start of the movement of the rotor to the time when the speed of the rotor reaches the target speed; and determining the target value of the starting time of the motor according to the time from the standstill of the rotor to the start of movement and the time from the start of movement of the rotor to the time when the speed of the rotor reaches the target speed.

In an embodiment, when the processor 302 executes the step of inputting the motor start control parameters in the sample set of the target motor into the target deep learning network model and obtaining the optimal motor start control parameters of the target motor according to the motor start prediction value output by the target deep learning network model, it may further perform: preprocessing the motor starting control parameters in the sample set of the target motor according to a preset searching method; inputting motor starting control parameters in a sample set of the preprocessed target motor into the target deep learning network model; obtaining an optimal motor starting predicted value from the motor starting predicted values output by the target deep learning network model; and acquiring a motor starting control parameter corresponding to the optimal motor starting predicted value, and taking the motor starting control parameter corresponding to the optimal motor starting predicted value as the optimal motor starting control parameter.

In an embodiment, before executing the randomly generating the motor start control parameter according to the preset value range of the motor start control parameter in the motor start model, and building the training set and the test set based on the randomly generated motor start control parameter, the processor 302 may further execute: and selecting the motor starting model.

In the embodiments of the motor configuration device and the readable storage medium provided in the present application, all technical features of the embodiments of the foregoing method are included, and the expanding and explaining contents of the specification are the same as the adaptability of the embodiments of the foregoing positioning method, and are not described again here.

Fig. 7 shows a schematic structural diagram of a motor start control system according to an embodiment of the present application. The motor start control system 400 comprises a main control end 401 and a motor drive board 402 which are connected, wherein the main control end 401 can be a PC end, a cloud computer and the like, and is used for generating test data, collecting test results, generating a test set, establishing a deep learning network model, constructing a trained model and generating motor start control parameters.

The motor driving board 402 is a driving board of a motor power supply and a motor peripheral circuit. The motor driving board 402 is provided with a motor driving chip 403, the motor driving chip 403 performs a start test on a motor start control parameter sent by the main control terminal 401, executes a motor drive control algorithm, controls the motor to rotate, collects a state parameter (such as voltage, current, time and the like) in a motor start process, and sends the state parameter to the main control terminal 401 for analysis, so that the main control terminal 401 determines an optimal motor start control parameter.

Embodiments of the present application also provide a computer program product comprising computer program code which, when run on a computer, causes the computer to perform the method as described in the various possible embodiments above.

Embodiments of the present application further provide a chip, which includes a memory and a processor, where the memory is used to store a program, and the processor is used to call and run the program from the memory, so that a device in which the chip is installed executes the method in the above various possible embodiments.

In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and a part which is not described in detail in a certain embodiment may refer to the above detailed description of the motor start control parameter optimization method, and is not described herein again.

The motor start control parameter optimization device provided in the embodiment of the present application and the motor start control parameter optimization method in the above embodiment belong to the same concept, and any method provided in the motor start control parameter optimization method embodiment may be operated in the motor start control parameter optimization, and the specific implementation process thereof is described in the motor start control parameter optimization method embodiment, and is not described herein again.

It should be noted that, for the method for optimizing motor start-up control parameters described in the embodiment of the present application, it may be understood by those skilled in the art that all or part of the process for implementing the method for optimizing motor start-up control parameters described in the embodiment of the present application may be implemented by controlling related hardware through a computer program, where the computer program may be stored in a computer-readable storage medium, such as a memory, and executed by at least one processor, and the process of implementing the method for optimizing motor start-up control parameters may be included in the process of execution. The storage medium may be a magnetic disk, an optical disk, a Read Only Memory (ROM), a Random Access Memory (RAM), or the like.

For the motor start control parameter optimization device in the embodiment of the present application, each functional module may be integrated in one processing chip, or each module may exist alone physically, or two or more modules are integrated in one module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode. The integrated module, if implemented in the form of a software functional module and sold or used as a stand-alone product, may also be stored in a computer readable storage medium, such as a read-only memory, a magnetic or optical disk, or the like.

The method, the device, the storage medium, the motor configuration device and the motor start control system for optimizing the motor start control parameters provided by the embodiments of the present application are described in detail above, specific examples are applied in the present application to explain the principle and the implementation of the present application, and the description of the above embodiments is only used to help understanding the method and the core idea of the present application; meanwhile, for those skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.

Claims

1. A motor starting control parameter optimization method is characterized by comprising the following steps:

randomly generating motor starting control parameters according to a preset value range of the motor starting control parameters in a motor starting model, wherein the motor starting control parameters comprise motor starting input parameters and motor starting target values, and the motor starting model is selected according to a current application scene;

generating online observation data through online monitoring;

establishing a training set and a test set according to the motor starting input parameters and the motor starting target values;

establishing a deep learning network model, training the deep learning network model by using the training set and a mean square error loss function, and verifying the trained model by using the test set and a mean absolute error method to obtain a final target deep learning network model;

and inputting motor starting control parameters in a sample set of a target motor into the target deep learning network model, and acquiring the optimal motor starting control parameters of the target motor according to a motor starting predicted value output by the target deep learning network model.

2. The method of optimizing motor start control parameters according to claim 1, wherein the establishing a training set and a test set based on the motor start input parameters and the motor start target values comprises:

3. The motor start control parameter optimization method of claim 1, wherein the motor start input parameters comprise at least:

4. The motor start-up control parameter optimization method according to claim 1, wherein the motor start-up target value includes at least any one of a motor start-up power target value and a motor start-up time target value.

5. The motor start control parameter optimization method according to claim 1, wherein when the motor start target value includes the motor start power target value, the online observation data includes real-time monitoring of a total current, a time from a standstill of a rotor to a start of movement, and a time from a start of movement of a rotor to a time at which a speed thereof reaches a target speed, the determining of the motor start target value based on the online observation data includes:

6. The motor start control parameter optimization method according to claim 4, wherein when the motor start target value includes the motor start time target value, the online observation data includes a time from a standstill of the rotor to a start of movement and a time from the start of movement of the rotor to a speed thereof to reach a target speed, the determining the motor start target value based on the online observation data includes:

7. The motor start control parameter optimization method according to claim 4, wherein when the motor start target value includes the motor start power target value and the motor start time target value, the online observation data includes real-time monitoring of a total current, a time from a standstill of a rotor to a start of movement, and a time from a start of movement of the rotor to a speed thereof reaching a target speed, the determining the motor start target value based on the online observation data includes:

8. The method for optimizing motor start control parameters according to claim 1, wherein the step of inputting the motor start control parameters in the sample set of the target motor into the target deep learning network model and obtaining the optimal motor start control parameters of the target motor according to the motor start prediction values output by the target deep learning network model comprises:

9. The method of claim 1, wherein before randomly generating motor start control parameters according to a preset value range of the motor start control parameters in the motor start model and building a training set and a testing set based on the randomly generated motor start control parameters, the method further comprises:

and selecting the motor starting model.

10. An electric machine start-up control parameter optimizing apparatus, comprising:

the system comprises a first establishing module, a second establishing module and a third establishing module, wherein the first establishing module is used for randomly generating motor starting control parameters according to a preset value range of the motor starting control parameters in a motor starting model, the motor starting control parameters comprise motor starting input parameters and motor starting target values, the motor starting model is selected according to a current application scene, online observation data are generated through online monitoring, the motor starting target values are determined according to the online observation data, and a training set and a testing set are established according to the motor starting input parameters and the motor starting target values;

the second establishing module is used for establishing a deep learning network model, training the deep learning network model by using the training set and a mean square error loss function, and verifying the trained model by using the test set and a mean absolute error method to obtain a final target deep learning network model;

11. A computer-readable storage medium, on which a computer program is stored, which, when executed on a computer, causes the computer to carry out the method of any one of claims 1 to 9.

12. A motor configuration device comprising a memory and a processor, characterized in that the processor executes the method according to any of claims 1 to 9 by invoking a computer program stored in the memory.

13. A motor starting control system comprises a main control end and a motor drive board which are connected, and is characterized in that the main control end is used for generating a training set and a testing set, establishing a deep learning network model and generating motor starting control parameters, and training the deep learning network model to obtain a final target deep learning network model;

the motor starting control parameters comprise motor starting input parameters and motor starting target values, and the motor starting model is selected according to the current application scene;

the final training method of the target deep learning network model comprises the following steps: training the deep learning network model by using the training set and a mean square error loss function, and verifying the trained model by using the test set and an average absolute error method to obtain a target deep learning network model;

the method for generating the training set and the test set comprises the following steps:

generating online observation data through online monitoring;