阅读说明：本技术 一种永磁同步电机输出扭矩估算方法及系统 (Output torque estimation method and system for permanent magnet synchronous motor ) 是由 李玮 刘超 梁海强 于 2019-01-31 设计创作，主要内容包括：本发明提供了一种永磁同步电机输出扭矩估算方法及系统,涉及汽车技术领域。该永磁同步电机输出扭矩估算方法,包括：在永磁同步电机处于预定义的低转速工况时,获取电机参数；将电机参数与当前扭矩命令,输入至一预先训练得到的径向基函数RBF神经网络,得到输出扭矩；对输出扭矩进行卡尔曼滤波,得到修正输出扭矩。通过引入RBF神经网络解决电机参数与输出扭矩的非线性关系,同时提出了一种获取训练数据的方法,保证了神经网络的计算精度,还引入卡尔曼滤波对输出扭矩进行处理,进一步提高精度,本方案不涉及对系统硬件的更改,因此不会增加系统成本。(The invention provides a method and a system for estimating output torque of a permanent magnet synchronous motor, and relates to the technical field of automobiles. The output torque estimation method of the permanent magnet synchronous motor comprises the following steps: when the permanent magnet synchronous motor is in a predefined low-rotation-speed working condition, motor parameters are obtained; inputting the motor parameters and the current torque command into a Radial Basis Function (RBF) neural network obtained by pre-training to obtain output torque; and performing Kalman filtering on the output torque to obtain a corrected output torque. The nonlinear relation between the motor parameters and the output torque is solved by introducing the RBF neural network, meanwhile, a method for acquiring training data is provided, the calculation precision of the neural network is guaranteed, Kalman filtering is introduced to process the output torque, the precision is further improved, and the scheme does not involve the change of system hardware, so that the system cost is not increased.)

1. A method for estimating an output torque of a permanent magnet synchronous motor of a vehicle, comprising:

when the permanent magnet synchronous motor is in a predefined low rotating speed working condition, obtaining motor parameters, wherein the motor parameters comprise: the current rotating speed of the motor, the current temperature of the motor, the current direct current bus voltage and the current working current of the motor;

inputting the motor parameters and the current torque command into a Radial Basis Function (RBF) neural network obtained by pre-training to obtain output torque;

and performing Kalman filtering on the output torque to obtain a corrected output torque.

2. The output torque estimation method of a permanent magnet synchronous motor according to claim 1, wherein the obtaining motor parameters includes:

collecting the motor parameters through a sensor;

and carrying out filtering processing on the motor parameters.

3. The pm synchronous motor output torque estimation method as claimed in claim 1, wherein prior to said obtaining motor parameters, said method further comprises:

acquiring training data of the RBF neural network;

and training the RBF neural network through the training data.

4. The output torque estimation method of a permanent magnet synchronous motor according to claim 3, wherein the acquiring training data includes:

obtaining a plurality of sets of training data by repeating the steps of:

placing a driving system provided with the permanent magnet synchronous motor in a test bench and connecting the driving system with an upper computer;

setting a torque command of the permanent magnet synchronous motor through an upper computer;

adjusting the power supply voltage of the driving system through the test bed;

the load is adjusted through the test bed, and the rotating speed of the permanent magnet synchronous motor is controlled to be in a predefined low rotating speed interval;

when the driving system runs to reach a preset stable condition, a torque command, the motor rotating speed, the direct current bus voltage, the motor temperature and the motor current are obtained through the upper computer, the motor output torque is obtained through the test bed, and a set of training data is obtained.

5. The output torque estimation method of a permanent magnet synchronous motor according to claim 1, wherein the kalman filtering the output torque includes:

and establishing a state equation and an observation equation, and correcting the output torque in an iterative mode.

6. An output torque estimation system of a permanent magnet synchronous motor, comprising:

the first obtaining module is used for obtaining motor parameters when the permanent magnet synchronous motor is in a predefined low rotating speed working condition, wherein the motor parameters comprise: the current rotating speed of the motor, the current temperature of the motor, the current direct current bus voltage and the current working current of the motor;

the processing module is used for inputting the motor parameters and the current torque command into a Radial Basis Function (RBF) neural network obtained by pre-training to obtain output torque;

and the filtering module is used for performing Kalman filtering on the output torque to obtain a corrected output torque.

7. The PMSM output torque estimation system of claim 6, wherein the first acquisition module includes:

the acquisition unit is used for acquiring the motor parameters through a sensor;

and the filtering unit is used for carrying out filtering processing on the motor parameters.

8. The pm synchronous motor output torque estimation system as claimed in claim 6, further comprising:

the second acquisition module is used for acquiring training data of the RBF neural network;

and the training module is used for training the RBF neural network through the training data.

Technical Field

The invention relates to the technical field of automobiles, in particular to a method and a system for estimating output torque of a permanent magnet synchronous motor.

Background

The pure electric vehicle drives wheels through a motor to realize vehicle running, a Permanent Magnet Synchronous Motor (PMSM) is generally applied along with the development of permanent magnet materials, power electronic technology, control theory, motor manufacturing and signal processing hardware, and the PMSM is the mainstream of a pure electric vehicle driving system at present due to the advantages of high efficiency, high output torque, high power density, good dynamic performance and the like. The safety and reliability are the basic requirements of the normal operation of the pure electric vehicle, and the realization of correct, effective and safe functions of a driving system (comprising a motor and a motor controller) in the vehicle is the premise of ensuring the safe operation of the vehicle. For a pure electric vehicle, the correct output of the torque of a driving system is the most basic premise of driving safety, and compared with a traditional fuel vehicle, the driving system of the pure electric vehicle relates to numerous high-voltage and low-voltage parts, and has larger potential failure risks, and the unexpected output of the torque of a driving motor is the most serious in the failure risks, so that the unexpected output of the torque of the driving system needs to be prevented at any time and in any state, so as to avoid accidents related to personal and vehicle safety.

At present, in the field of pure electric vehicles, an estimation method for the output torque of a permanent magnet synchronous motor is mature, the estimation accuracy is good in the aspect of medium and high rotating speeds, but the estimation accuracy is still to be improved under the working condition of low rotating speed.

Disclosure of Invention

The embodiment of the invention provides a method and a system for estimating output torque of a permanent magnet synchronous motor, which are used for solving the problem that perturbation of the permanent magnet synchronous motor under a low-rotating-speed working condition affects the estimation precision of the output torque.

In order to solve the above technical problem, an embodiment of the present invention provides a method for estimating an output torque of a permanent magnet synchronous motor of a vehicle, including:

when the permanent magnet synchronous motor is in a predefined low rotating speed working condition, obtaining motor parameters, wherein the motor parameters comprise: the current rotating speed of the motor, the current temperature of the motor, the current direct current bus voltage and the current working current of the motor;

inputting the motor parameters and the current torque command into a Radial Basis Function (RBF) neural network obtained by pre-training to obtain output torque;

and performing Kalman filtering on the output torque to obtain a corrected output torque.

Further, the acquiring the motor parameter includes:

collecting the motor parameters through a sensor;

and carrying out filtering processing on the motor parameters.

Further, before the obtaining of the motor parameter, the method further comprises:

acquiring training data of the RBF neural network;

and training the RBF neural network through the training data.

Further, the acquiring training data comprises:

obtaining a plurality of sets of training data by repeating the steps of:

placing a driving system provided with the permanent magnet synchronous motor in a test bench and connecting the driving system with an upper computer;

setting a torque command of the permanent magnet synchronous motor through an upper computer;

adjusting the power supply voltage of the driving system through the test bed;

the load is adjusted through the test bed, and the rotating speed of the permanent magnet synchronous motor is controlled to be in a predefined low rotating speed interval;

when the driving system runs to reach a preset stable condition, a torque command, the motor rotating speed, the direct current bus voltage, the motor temperature and the motor current are obtained through the upper computer, the motor output torque is obtained through the test bed, and a set of training data is obtained.

Further, the kalman filtering the output torque includes:

and establishing a state equation and an observation equation, and correcting the output torque in an iterative mode.

The embodiment of the present invention further provides a system for estimating an output torque of a permanent magnet synchronous motor, including:

the first obtaining module is used for obtaining motor parameters when the permanent magnet synchronous motor is in a predefined low rotating speed working condition, wherein the motor parameters comprise: the current rotating speed of the motor, the current temperature of the motor, the current direct current bus voltage and the current working current of the motor;

the processing module is used for inputting the motor parameters and the current torque command into a Radial Basis Function (RBF) neural network obtained by pre-training to obtain output torque;

and the filtering module is used for performing Kalman filtering on the output torque to obtain a corrected output torque.

Further, the first obtaining module comprises:

the acquisition unit is used for acquiring the motor parameters through a sensor;

and the filtering unit is used for carrying out filtering processing on the motor parameters.

Further, the system further comprises:

the second acquisition module is used for acquiring training data of the RBF neural network;

and the training module is used for training the RBF neural network through the training data.

The invention has the beneficial effects that:

according to the scheme, the nonlinear relation between the motor parameters and the output torque is solved by introducing the RBF neural network, meanwhile, a method for acquiring training data is provided, the calculation precision of the neural network is guaranteed, Kalman filtering is introduced to process the output torque, the precision is further improved, and the scheme does not involve the change of system hardware, so that the system cost is not increased.

Drawings

FIG. 1 is a flow chart illustrating a method for estimating output torque of a permanent magnet synchronous motor according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of an RBF neural network according to an embodiment of the present invention;

FIG. 3 is a flow chart illustrating an embodiment of obtaining RBF neural network training data;

fig. 4 is a block diagram of an output torque estimation system of a permanent magnet synchronous motor according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in detail with reference to the accompanying drawings and specific embodiments.

The invention provides a method and a system for estimating output torque of a permanent magnet synchronous motor, aiming at the problem that perturbation of motor parameters of the permanent magnet synchronous motor under a low-rotating-speed working condition can influence the estimation precision of the output torque.

As shown in fig. 1, an embodiment of the present invention provides a method for estimating an output torque of a permanent magnet synchronous motor, including:

step 11, when the permanent magnet synchronous motor is in a predefined low-rotation-speed working condition, obtaining motor parameters, wherein the motor parameters comprise: the current rotating speed of the motor, the current temperature of the motor, the current direct current bus voltage and the current working current of the motor;

it should be noted that the predefined low-rotation-speed operating condition refers to a region where the rotation speed of the motor is in a low-rotation-speed interval, and is usually not greater than a certain predetermined rotation speed, where the predetermined rotation speed may be set according to a specific motor model, for example, 1000rpm may be used, and also 500rpm may be used.

Step 12, inputting the motor parameters and the current torque command into a Radial Basis Function (RBF) neural network obtained by pre-training to obtain output torque;

it should be noted that, the output torque of the motor under the low rotation speed condition has a high correlation with the motor working current, and is influenced by the motor temperature, the motor rotation speed and the dc bus voltage, perturbation of the motor parameter is an important factor influencing the torque estimation accuracy, a complex nonlinear relation exists between the output torque and the motor parameter, which cannot be described by an accurate mathematical expression, the neural network method has a nonlinear basic characteristic, and has natural advantages for solving the nonlinear problem, while the RBF neural network, as a feedforward neural network with excellent performance, can approach any nonlinear function with any accuracy, and has a compact topology structure, global approximation capability, and no local optimization problem of the BP network And (4) the external environment parameters are linked with the output torque of the motor, so that the current output torque of the driving motor is estimated.

And step 13, performing Kalman filtering on the output torque to obtain a corrected output torque.

It should be noted that, in order to further reduce the influence of the uncertainty of the system on the estimation accuracy, the embodiment of the present invention introduces kalman filtering, processes the output torque obtained through the RBF neural network, and finally obtains the corrected output torque under the low-speed working condition, thereby further improving the reliability of the estimation of the output torque of the motor under the low-speed working condition.

Specifically, the acquiring of the motor parameter in step 11 includes:

collecting the motor parameters through a sensor;

and carrying out filtering processing on the motor parameters.

It should be noted that due to the operating characteristics of the pure electric vehicle, a large amount of high-frequency interference must be included in the motor parameters acquired by the sensor, and therefore, in order to prevent the unexpected interference from damaging the calculation accuracy of the RBF neural network and to ensure the stability of the calculation result, the acquired motor parameters need to be filtered in advance.

Filtering the current rotating speed of the motor

Preferably, rolling filtering may be adopted to filter the motor rotation speed, and the specific expression is as follows:

S(n)＝K_sS_int(n)+(1-K_s)S_int(n-1)，K_s∈[0，1]

wherein S (n) represents the filtered motor rotor obtained in the control periodSpeed; k_sIs a weight coefficient; s_int(n) the motor rotating speed acquired in the control period; s_intAnd (n-1) is the motor rotating speed acquired in the previous control period. Under normal conditions, the change of the rotating speed of the motor is slow, so that the rolling filtering mode is suitable.

Filtering the current temperature of the motor

The motor temperature is obtained by analyzing a voltage signal fed back by a temperature-sensitive resistor embedded in a motor stator winding, a large amount of high-frequency interference exists in the voltage signal fed back by the temperature-sensitive resistor in consideration of the actual working environment of the motor, a second-order low-pass filtering mode is adopted for the motor temperature, the high-frequency interference is filtered, and the specific expression is as follows:

T(n)＝f_L(n)-f_L(n-2)

wherein, T (n) represents the motor temperature obtained by low-pass filtering in the control period; f. of_L(n)＝T_o(n)K_La-K_Lbf_L(n-1)-K_Lcf_L(n-2)，T_o(n) represents the motor temperature value collected in the control cycle, K_La、K_LbAnd K_LcRepresenting the filter coefficients, which are used to adjust parameters such as the low pass filter cut-off frequency.

3, current DC bus voltage filtering

Direct current busbar voltage is likewise obtained through sensor feedback signal analysis, and similar to motor temperature sensor feedback signal, there is high-frequency disturbance all, for this reason adopts the second order low pass filtering mode to current direct current busbar voltage, filters high-frequency disturbance, and the concrete expression is:

U(n)＝f_L(n)-f_L(n-2)

wherein, U (n) represents the DC bus voltage obtained by low-pass filtering in the control period; f. of_L(n)＝T_o(n)K_La-K_Lbf_L(n-1)-K_Lcf_L(n-2)，U_o(n) represents the DC bus voltage collected in the control period, K_La、K_LbAnd K_LcRepresenting the filter coefficients, which are used to adjust parameters such as the low pass filter cut-off frequency.

4, filtering the current working current of the motor

The current working current of the motor refers to the dq axis current, and the dq axis current needs to be obtained through Clark conversion and Park conversion of the actual U \ V \ W three-phase current of the motor, so that the current working current filtering of the motor is the U \ V \ W three-phase current filtering.

An average value method is adopted for three-phase current filtering, and specifically comprises the following steps: in a control period, sampling is carried out on three-phase currents of U \ V \ W for K times by using a current sensor, a maximum sampling value and a minimum sampling value are removed, then average filtering is carried out on the remaining sampling values of K-2 times to obtain a filtered signal, and finally the current value of the U \ V \ W phase is analyzed by using the weighted average signal. Obtaining the current dq axis current of the motor by using the filtered three-phase current through coordinate transformation, and defining the current as i_d(n) and i_q(n)。

As shown in fig. 2, the RBF neural network adopted in the embodiment of the present invention is divided into three layers, an input layer, a hidden layer and an output layer, wherein the number of input is 6, and the input is the current torque command T of the motor_cmdThe current rotating speed S of the motor, the current temperature T of the motor, the current direct current bus voltage U and the current working current i of the motor_dAnd i_qThe number of neurons in the hidden layer is 13, and the output quantity is the estimated output torque T of the motor_intThe specific expression is as follows:

where x is the input vector, i.e. x ═ T_cmdS T U i_di_q]Wherein T is_cmdRepresenting the current torque command of the motor, S representing the current rotating speed of the motor, T representing the temperature of the motor, U representing the direct current bus voltage, i_dAnd i_qRepresenting the current working current of the motor; y (x, w) is the net output, i.e. the calculated motor output torque T_int；w_iIs a weight; l is the number of hidden layer neurons, and l is 13; c. C_iIs a central vector; i x-c_i| is the distance from the input vector to the node center (center vector);is a radial basis function, here taken as a gaussian radial basis function.

It should be noted that, the quality of the calculation accuracy of the RBF neural network depends on the data used for training the RBF neural network, and if the data used for training the RBF neural network is real, accurate and reliable, the RBF neural network can obtain excellent calculation accuracy through training, so before step 11, the method further includes:

acquiring training data of the RBF neural network;

and training the RBF neural network through the training data.

It should be noted that there are many mature methods for training the RBF neural network, which are not described herein, and in particular, the embodiment of the present invention provides a method for acquiring training data.

Specifically, the acquiring training data includes:

by repeating the following steps, a plurality of sets of training data are obtained, as shown in fig. 3, the method for obtaining a set of training data is:

step 31, placing a driving system provided with the permanent magnet synchronous motor in a test bench and connecting the driving system with an upper computer;

step 32, setting a torque command of the permanent magnet synchronous motor through an upper computer;

step 33, adjusting the power supply voltage of the driving system through the test bed;

step 34, adjusting a load through a test bed, and controlling the rotating speed of the permanent magnet synchronous motor to be in a predefined low rotating speed interval;

and step 35, when the operation of the driving system reaches a preset stable condition, acquiring a torque command, the rotating speed of the motor, the voltage of a direct current bus, the temperature of the motor and the current of the motor through the upper computer, and acquiring the output torque of the motor through the test bed to acquire a group of training data.

The test bench can obtain the output power and the output torque of the driving motor in real time, and can adjust the power supply voltage of the driving system; the stable condition may include: the variation range of the motor rotating speed, the output torque and the motor temperature in unit time does not exceed the corresponding range threshold, and the unit time and the preset range threshold can be set according to empirical values.

And training the RBF neural network by using the obtained multiple groups of training data, so that the RBF neural network has accurate motor output torque estimation performance.

Considering that filtering processing is performed on input signals input into the RBF neural network, but influence of interference on calculation results of the RBF neural network cannot be completely eliminated, and in order to further reduce influence of unexpected disturbance on output torque, the embodiment of the invention combines the characteristics of Kalman filtering and filters the output torque output by the RBF neural network by establishing a state equation and an observation equation, thereby further improving the reliability of low-rotation-speed working condition motor output torque estimation.

Kalman filtering the output torque, comprising:

and establishing a state equation and an observation equation, and correcting the output torque in an iterative mode.

Specifically, the state equation of the filter system is T_q(n)＝T_q(n-1) + W (n-1), wherein T_q(n) is a one-dimensional variable output torque, namely an estimated value of the output torque of the driving motor in the nth control period obtained after Kalman filtering processing; w represents process noise, the variance of the process noise is Q, and Q can be a small fixed value in practical application;

the observation equation of the filter system is T_int(n)＝T_q(n) + V (n) where T_int(n) the motor output torque of the nth control period calculated by the RBF neural network is represented; v represents the observed noise of the neural network, with variance R, which can also be a small fixed value in practical applications.

And the Kalman filtering is used for predicting the actual output torque of the motor in the nth control period by utilizing the motor output torque calculated in the nth-1 control period through the RBF neural network.

The estimated deviation of the filter system is defined as P (n | n-1), where P (n | n-1) denotes the deviation of the (n-1) th control cycle, and P (n | n-1) + Q, from which the kalman gain K, K ═ P (n | n-1)/[ P (n | n-1) + R ] can be obtained.

In summary, the kalman filter expression can be obtained:

T_q(n)＝T_int(n-1)+K[T_int(n)-T_int(n-1)]

wherein, T_q(n) the motor output torque obtained after the nth control period is subjected to Kalman filtering, namely the final required correction output torque is represented; t is_int(n) the motor output torque calculated by the RBF neural network in the nth control period, T_intAnd (n-1) represents the motor output torque calculated by the RBF neural network in the (n-1) th control cycle.

Completion of T_qAfter the calculation of (n), P needs to be updated, that is, the deviation P (n) of the nth control period is calculated, and the expression is as follows:

P(n)＝(1-K)P(n|n-1)

after the deviation P (n) is calculated, Kalman filtering is performed in the next period in the form, and the reliability of the output torque estimation of the permanent magnet synchronous motor of the pure electric vehicle under the working condition of low rotating speed is improved by continuously iterating in the Kalman filtering iteration mode.

As shown in fig. 4, an embodiment of the present invention further provides a system for estimating output torque of a permanent magnet synchronous motor, including:

the first obtaining module 41 is configured to obtain a motor parameter when the permanent magnet synchronous motor is in a predefined low rotation speed condition, where the motor parameter includes: the current rotating speed of the motor, the current temperature of the motor, the current direct current bus voltage and the current working current of the motor;

the processing module 42 is configured to input the motor parameter and the current torque command to a radial basis function RBF neural network obtained through pre-training to obtain an output torque;

and a filtering module 43, configured to perform kalman filtering on the output torque to obtain a corrected output torque.

The first obtaining module comprises:

the acquisition unit is used for acquiring the motor parameters through a sensor;

and the filtering unit is used for carrying out filtering processing on the motor parameters.

The output torque estimation system of the permanent magnet synchronous motor further comprises:

the second acquisition module is used for acquiring training data of the RBF neural network;

and the training module is used for training the RBF neural network through the training data.

It should be noted that, in the embodiment of the present invention, the operating state of the motor and the external environment parameters are used as inputs, including the current torque command of the driving motor, the motor speed, the motor temperature, the dc bus voltage, and the current operating current of the motor, and the output torque of the motor in the current state is estimated by using the RBF neural network. Considering that ideal training data is an important premise for ensuring the calculation accuracy of the RBF, aiming at the RBF neural network provided by the invention, the invention also provides a training data acquisition method, and the calculation accuracy can be ensured by training the RBF neural network by using the data obtained by the method. The RBF neural network is trained according to ideal data in advance, so that the RBF neural network has the advantage of high speed in the actual application process, and cannot influence the real-time performance in the vehicle control process. In addition, the method for estimating the output torque of the permanent magnet synchronous motor of the pure electric vehicle is a method with good engineering realizability, has the advantages of reliability, effectiveness, high accuracy, easiness in implementation and the like, and does not involve the change of system hardware, namely the system cost is not increased.

While the preferred embodiments of the present invention have been described, it will be understood by those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the invention as defined in the following claims.