CN114744938B - A full parameter observer and full parameter identification method based on Kalman filtering - Google Patents

Abstract

The invention discloses a Kalman filtering-based full-parameter observer and a full-parameter identification method, which belong to the field of parameter identification of permanent magnet synchronous motors and comprise the following steps: the system comprises a parameter prior prediction module, a parameter posterior estimation module, a state prior prediction module and a state posterior estimation module; the parameter priori prediction module is used for performing priori prediction on the parameters of the current period to obtain the parameter priori predicted value of the current periodThe state priori prediction module is used for performing priori prediction on the state of the current period to obtain a state priori predicted value of the current periodThe parameter posterior estimation module is used for performing posterior estimation on the parameters of the current period to obtain the posterior estimation value of the parameters of the current periodThe state posterior estimation module is used for performing posterior estimation on the state of the current period to obtain a state posterior estimation value of the current periodThe invention can realize accurate identification of parameters in the running process of the permanent magnet synchronous motor under the full working condition.

Description

A full parameter observer and full parameter identification method based on Kalman filtering

Technical Field

The present invention belongs to the field of parameter identification of permanent magnet synchronous motors, and more specifically, relates to a full-parameter observer based on Kalman filtering and a full-parameter identification method.

Background technique

At present, various industries are undergoing a transformation towards electrification and intelligence. Among them, the electrification of household appliances, more electric aircraft and electric vehicles is rapidly becoming popular, and motors, as the key power parts of electrification systems, naturally face more challenges and opportunities. Compared with traditional industrial scene motors, motors in such application scenarios must have a wider speed operating range and stronger overload capacity to meet the speed regulation and traction requirements of the power system, and due to space limitations, they must have higher power density and efficiency to reduce the volume share of the cooling system. Among the many types of motors, permanent magnet synchronous motors have become the first choice for the main drive motors of electric vehicles because of their strong overload capacity and wide range of weak magnetic field capabilities. Moreover, in terms of electric drive control, it is necessary not only to achieve high-efficiency operation, but also to achieve precise speed and torque control within the entire operating speed range. In order to meet the above application requirements, it is necessary to obtain accurate parameters under the operating state of the motor to achieve precise and intelligent control of the system. However, the existing methods and mathematical models have the following disadvantages:

The traditional motor drive control system and parameter identification system are both based on the mathematical model of the linear permanent magnet synchronous motor, which is expressed as follows:

Among them, _Rs represents the stator winding of the motor, _ψf represents the permanent magnet flux, _Ld and _Lq represent the d-axis and q-axis inductances respectively, _ud and _uq represent the d-axis and q-axis voltages respectively, id and _iq represent the d _-axis and q-axis currents respectively, and _ωe represents the motor fundamental frequency.

The above simplified mathematical model can facilitate the design of the controller, but in fact, the motor is a highly coupled and strongly nonlinear system. When the motor runs at a large current, the model cannot reflect the actual electromagnetic conditions inside the motor. Accordingly, the accuracy of the parameter identification results cannot be guaranteed.

The traditional parameter identification algorithm based on high-frequency injection will inject high-frequency disturbance voltage into the voltage command output by the motor current loop, and directly ignore the ω _e L _d i _d and ω _e L _q i _q items in the parameter identification process based on the assumption that the frequency of the injected voltage signal ω _h is much greater than the fundamental frequency ω _e of the motor operation. Due to the limitation of the inverter switching frequency, the frequency of the injected high-frequency signal cannot be increased infinitely, so this scheme can only realize parameter identification under zero speed and low speed conditions. This assumption does not hold true when the motor runs at high speed. In addition, since the embedded permanent magnet synchronous motor has the characteristic of L _d << L _q , ignoring the ω _e L _d i _d and ω _e L _q i _q items will also cause large errors in the observation of inductance, thereby affecting the accuracy of parameter identification.

Summary of the invention

In view of the defects of the prior art and the need for improvement, the present invention provides a full-parameter observer and a full-parameter identification method based on Kalman filtering, the purpose of which is to achieve accurate identification of parameters during the operation of a permanent magnet synchronous motor under all operating conditions.

To achieve the above object, according to one aspect of the present invention, a full parameter observer based on Kalman filtering is provided, which is used for parameter identification of a permanent magnet synchronous motor. During the operation of the permanent magnet synchronous motor, the d-axis and q-axis command voltages output by the current loop are respectively injected with small signal voltages u _dh and u _qh with a frequency of ω _h ; the full parameter observer includes: a parameter extended Kalman filter and a state extended Kalman filter;

The parameter extended Kalman filter includes: a parameter priori prediction module and a parameter a posteriori estimation module; the state extended Kalman filter includes: a state priori prediction module and a state a posteriori estimation module;

A parameter a priori prediction module, whose first input terminal is connected to the output terminal of the parameter posterior estimation module, is used to perform a priori prediction on the parameters of the current cycle to obtain a priori prediction value θ _k ⁻ of the parameters of the current cycle;

The state prior prediction module has a first input terminal connected to the output terminal of the state posterior estimation, and a second input terminal connected to the output terminal of the parameter prior prediction module, which is used to perform a priori prediction on the state of the current cycle to obtain the a priori prediction value of the state of the current cycle.

A parameter posterior estimation module, whose first input end is connected to the output end of the parameter a priori prediction module, and whose second input end is connected to the output end of the state a priori prediction module, is used to perform a posteriori estimation on the parameters of the current cycle to obtain the a posteriori estimation value of the parameters of the current cycle

A state posterior estimation module, whose first input end is connected to the output end of the state a priori prediction module, and whose second input end is connected to the output end of the parameter a priori prediction module, is used to perform a posteriori estimation on the state of the current cycle to obtain a posteriori estimation value of the state of the current cycle.

in, represents the state posterior estimation value of the previous cycle, U _k-1 represents the small signal voltage injected in the previous cycle; T _s represents the sampling time interval, ω _e represents the motor fundamental frequency; /> _Rs represents the stator winding resistance of the motor; Expressed as the inverse matrix of the inductance matrix,,/> and They represent the d-axis incremental self-inductance and the q-axis incremental self-inductance respectively,/> represents the incremental mutual inductance of d and q axes; θ＝[R _s T _dd T _dq T _qq ] ^T represents the parameter to be observed, /> f ^* () represents the discrete state equation of the motor.

Furthermore, the calculation expression of the parameter prior prediction module for signal processing is:

in, Represents the posterior estimate of the parameters of the previous cycle; /> Represents the parameter prior prediction covariance matrix of the current period,/> represents the covariance matrix of the posterior estimation of the parameters in the previous period, and _Qp represents the system noise covariance matrix in the prior prediction of the parameters.

Furthermore, the calculation expression of the state prior prediction model for signal processing is:

in, Represents the parameter prior estimation covariance matrix of the current period,/> represents the state posterior estimation covariance matrix of the previous cycle, Q _x represents the system noise covariance matrix in the state prior prediction; F _x,k-1 represents the value of the derivative of the discrete state equation f ^* () with respect to the motor state X* in the previous cycle.

Furthermore, the calculation expression of the signal processing performed by the parameter posterior estimation module is:

in, represents the output of the motor in the current cycle, h ^* () represents the discrete output equation of the motor; /> represents the posterior estimation covariance matrix of the parameters in the current cycle; K _θ,k represents the Kalman gain of the parameters in the current cycle; H _θ,k represents the value of the derivative of the discrete output equation with respect to the motor parameter θ in the current cycle; M represents the measurement noise covariance matrix of the system.

Furthermore, the calculation expression of the state posterior estimation module for signal processing is:

in, represents the posterior estimation covariance matrix of the state of the current cycle, K _x,k represents the Kalman gain of the state of the current cycle, and H _x,k represents the value of the derivative of the discrete output equation with respect to the motor state X* in the current cycle.

According to another aspect of the present invention, a method for full parameter identification of a permanent magnet synchronous motor based on the above-mentioned full parameter observer is provided, comprising:

Small signal disturbance voltage injection step: during the operation of the motor, a d-axis small signal disturbance voltage u _dh and a q-axis small signal disturbance voltage u _qh with a frequency of ω _h are respectively injected into the d-axis and q-axis command voltages output by the current loop of the motor;

The full parameter identification steps include:

(S1) sampling the three-phase current of the motor in the running state, and transforming it to a dq-axis rotating coordinate system whose synchronous speed is the motor fundamental frequency ω _e , to obtain the d-axis current i _d and the q-axis current i _q ;

(S2) extracting the components with frequency _ωh from the _d -axis current id and the q-axis current _iq respectively, and obtaining the d-axis small-signal disturbance current _idh and the q-axis small-signal disturbance current _iqh under the excitation of the d-axis small-signal disturbance voltage _udh and the q-axis small-signal disturbance voltage _uqh ;

(S3) Input the d-axis small signal disturbance current i _dh and the q-axis small signal disturbance current i _qh into the full parameter observer based on Kalman filtering, and obtain the parameter posterior estimation value θ = [R _s T _dd T _dq T _qq ] ^T output by the full parameter observer based on Kalman filtering to calculate the motor stator winding resistance R _s and the d-axis incremental self-inductance Q-axis incremental self-inductance/> And the incremental mutual inductance of d and q axes

(S4) Based on the parameters calculated in (S3), the d-axis magnetic flux ψ _d0 and the q-axis magnetic flux ψ _q0 are calculated.

Further, in step (S4), the calculation expressions for calculating the d-axis flux ψ _d0 and the q-axis flux ψ _q0 are:

Among them, u _d0 and u _q0 represent the d-axis fundamental frequency signal voltage and the q-axis fundamental frequency signal voltage respectively, and i _d0 and i _q0 represent the d-axis fundamental frequency signal current and the q-axis fundamental frequency signal current respectively.

Furthermore, between steps (S1) and (S2), the following steps are further included:

Filter out the components with a frequency of ω _h in the d-axis current i _d and the q-axis current i _q to obtain the d-axis direct current i _d0 and the q-axis direct current i _q0 when the motor is running;

The d-axis DC current i _d0 and the q-axis DC current i _q0 are input into the current loop.

In general, through the above technical scheme conceived by the present invention, the discretized motor state model f ^* () used fully considers the high-coupling and strong nonlinear characteristics of the permanent magnet synchronous motor, as well as the influence of the injected signal on the motor state when a high-frequency disturbance voltage signal is injected. Therefore, the motor state model f ^* () can more accurately reflect the state of the motor during operation, is not limited by the motor speed, and can achieve accurate observation of the motor parameters under all working conditions; in addition, in the present invention, two Kalman filters are designed for observing the motor state and motor parameters respectively, and the a priori prediction results of the two Kalman filters will be called by each other. Under the mutual promotion of state observation and parameter observation, the accuracy of parameter observation can be further improved, thereby improving the accuracy of subsequent motor parameter identification.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of a full parameter observer based on Kalman filtering provided by an embodiment of the present invention;

FIG2 is a schematic diagram of a control system of a permanent magnet synchronous motor provided in an embodiment of the present invention;

FIG3 is a schematic diagram of a method for identifying all parameters of a permanent magnet synchronous motor provided in an embodiment of the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

In the present invention, the terms "first", "second", etc. (if any) in the present invention and the drawings are used to distinguish similar objects but not necessarily to describe a specific order or sequence.

In order to solve the technical problem that the existing parameter identification method cannot accurately identify the parameters of the permanent magnet synchronous motor under all working conditions, the present invention provides a full parameter observer and full parameter identification method based on Kalman filtering, the overall idea of which is: when establishing the state equation of the motor, the high coupling and strong nonlinear characteristics of the permanent magnet synchronous motor, as well as the influence of the injected signal on the motor state when injecting a high-frequency disturbance voltage signal, are fully considered; on the basis of the established state equation, combined with the characteristics of the Kalman filter, the observation variables are designed accordingly to obtain an observer that can accurately identify the motor parameters. Based on the designed observer, the accurate identification of the parameters of the permanent magnet synchronous motor under all working conditions is achieved.

Before explaining the technical solution of the present invention in detail, the relevant principles are analyzed as follows.

The mathematical model of the linear permanent magnet synchronous motor is shown in formula (1):

Among them, _Rs represents the stator winding of the motor, _ψf represents the permanent magnet flux, _Ld and _Lq represent the d-axis and q-axis inductances respectively, _ud and _uq represent the d-axis and q-axis voltages respectively, id and _iq represent the d _-axis and q-axis currents respectively, and _ωe represents the motor fundamental frequency.

The mathematical model of the permanent magnet synchronous motor shown in formula (1) is converted to the synchronous rotating coordinate system, as shown in formula (2):

Among them, ψ _d and ψ _q represent the d-axis flux and q-axis flux, respectively; considering the case of injected small-signal disturbance voltage, the d- and q-axis voltages u _d and u _q can be expressed as the sum of the fundamental frequency signal voltage and the small-signal disturbance voltage, that is, u _d ＝u _d0 +u _dh , u _q ＝u _q0 +u _qh , u _d0 and u _q0 represent the d- and q-axis fundamental frequency signal voltages, respectively, and u _dh and u _qh represent the injected d- and q-axis small-signal disturbance voltages, respectively; similarly, the d- and q-axis currents i _d and i _q can be expressed as the sum of the fundamental frequency current signal and the small-signal disturbance current, that is, i _d ＝i _d0 +i _dh , i _q ＝i _q0 +i _qh , i _d0 and i _q0 represent the d- and q-axis fundamental frequency current signals, respectively, and i _dh and i _qh represent the d- and q-axis small-signal disturbance currents under the excitation of the small-signal disturbance voltage, respectively.

Considering that the permanent magnet synchronous motor is a highly coupled and strongly nonlinear system, the d-axis and q-axis flux ψ _d and ψ _q mainly depend on the permanent magnet flux ψ _f and the d-axis and q-axis currents i _d and i _q , so the flux can be considered to be a function of the d-axis and q-axis currents. Taylor expansion of the motor d-axis and q-axis flux ψ _d and ψ _q yields:

Wherein, a _dd , a _dq , a _qd , a _qq are the derivatives of the d-axis flux and the q-axis flux at the points i _d0 and i _q0 to the d-axis current and the q-axis current, respectively, which are called incremental inductance; HOT represents a high-order term. The amplitude of the small signal disturbance current excited by the disturbance voltage signal injected by the present invention is small (preferably less than 5% of the rated current of the motor). Since the amplitude of the disturbance current is small, the high-order term in formula (3) can be ignored;

The incremental inductance in formula (3) can be expressed by formula (4):

in, and/> They represent the d-axis incremental self-inductance and the q-axis incremental self-inductance respectively,/> and/> represents the incremental mutual inductance of the d and q axes; based on the magnetic circuit equivalence theorem, the two incremental mutual inductances are equal, that is,/>

By taking the derivative of formula (3), we can get:

Based on formulas (2) to (5), the equation of the permanent magnet synchronous motor under small signal can be expressed as (6):

By separating the large fundamental frequency voltage and current signals and the small disturbance signals through filtering and other means, the equations of the large signal model (7) and the small signal model (8) can be obtained respectively.

Since the above formulas (7) and (8) take into account the influence of the injected small signal disturbance voltage on the motor state, the motor model represented by the above formulas (7) and (8) can more accurately describe the internal state of the permanent magnet synchronous motor.

The present invention designs a full parameter observer based on the model shown in the above formulas (7) and (8), and the specific process is as follows:

By transforming formula (7), the state equation of the permanent magnet synchronous motor is obtained as follows:

Where X represents the state variable, represents the first-order derivative; Y represents the motor output, and I represents the unit matrix; since the state equation contains both the inductance matrix L and the inverse matrix L ^-1 of the inductance, it is difficult to design the subsequent observer; in order to facilitate the design of the observer, the present invention further rewrites the above formula (9) to form a form that is convenient for designing the observer. Specifically, by defining a new state variable X*=LX, a new state equation can be obtained as shown in formula (10):

In the state equation shown in formula (10), the inductance matrix L is eliminated, and only the inverse matrix L ^-1 of the inductance matrix exists.

Define the observed variable matrix as θ = [R _s T _dd T _dq T _qq ] ^T , θ includes some resistance parameters and inductance parameters to be identified in the permanent magnet synchronous motor, which can be regarded as an extended state variable; since the resistance parameters and inductance parameters in the motor change slowly over time, it can be considered that dθ/dt≈0. Based on this, the state equation after the extended variables can be obtained as:

in, represents the first-order derivative of parameter θ,/> represents the first-order derivative of the state variable X*; f represents the state equation of the state variable X*, and h represents the output equation of the output variable Y.

The state equation shown in the above formula (11) is discretized to obtain the following discretized state equation:

Where k represents the data point number in the discrete sequence, _Ts represents the sampling time interval, and f* and h* are the discretizations of f and h, respectively.

According to the discretized state equation shown in (12), in one embodiment of the present invention, a full parameter observer based on Kalman filtering is constructed. The observer is shown in FIG1 and includes two Kalman filters, namely, a parameter extended Kalman filter and a state extended Kalman filter;

The parameter extended Kalman filter includes: a parameter prior prediction module and a parameter posterior estimation module; the state extended Kalman filter includes: a state prior prediction module and a state posterior estimation module; as shown in Figure 1, the functions of each module, that is, the connection relationship between the modules is:

The parameter a priori prediction module has a first input terminal connected to the output terminal of the parameter posterior estimation module, which is used to make a priori prediction of the parameters of the current cycle to obtain the a priori prediction value of the parameters of the current cycle.

The state prior prediction module has a first input terminal connected to the output terminal of the state posterior estimation, and a second input terminal connected to the output terminal of the parameter prior prediction module, which is used to perform a priori prediction on the state of the current cycle to obtain the a priori prediction value of the state of the current cycle.

A parameter posterior estimation module, whose first input end is connected to the output end of the parameter a priori prediction module, and whose second input end is connected to the output end of the state a priori prediction module, is used to perform a posteriori estimation on the parameters of the current cycle to obtain the a posteriori estimation value of the parameters of the current cycle

A state posterior estimation module, whose first input end is connected to the output end of the state a priori prediction module, and whose second input end is connected to the output end of the parameter a priori prediction module, is used to perform a posteriori estimation on the state of the current cycle to obtain a posteriori estimation value of the state of the current cycle.

in, represents the state posterior estimation value of the previous cycle, U _k-1 represents the small signal voltage injected in the previous cycle; f ^* () represents the discrete state equation of the motor.

In the observer shown in Figure 1, in the same observation cycle, the state prior prediction depends on the result of the parameter prior prediction, the parameter posterior estimation depends on the result of the state prior prediction, and the state posterior estimation also depends on the result of the parameter prior prediction. The extended Kalman filtering on the state and the extended Kalman filtering on the parameter are performed synchronously. Therefore, the full-parameter observer based on Kalman filtering provided in this embodiment is a full-parameter observer based on parallel extended Kalman filtering.

In this embodiment, the calculation expressions for signal processing performed by each module are specifically as follows:

Parameter prior prediction:

in, represents the posterior estimate of the parameters of the previous cycle,/> is the prior prediction value of the parameters of this cycle;/> Represents the parameter prior prediction covariance matrix of the current period,/> represents the covariance matrix of the posterior estimation of the parameters in the previous cycle, _Qp represents the system noise covariance matrix in the prior prediction of the parameters;

State prior estimate:

in, Represents the parameter prior estimation covariance matrix of the current period,/> represents the state posterior estimation covariance matrix of the previous cycle, Q _x represents the system noise covariance matrix in the state prior prediction; F _x,k-1 represents the value of the derivative of the discrete state equation f ^* () with respect to the motor state X* in the previous cycle;

Posterior estimate of parameters:

in, represents the output of the motor in the current cycle, and h ^* () represents the discrete output equation of the motor; /> represents the posterior estimation covariance matrix of the parameters in the current cycle; K _θ,k represents the Kalman gain of the parameters in the current cycle; H _θ,k represents the value of the derivative of the discrete output equation with respect to the motor parameter θ in the current cycle; M represents the measurement noise covariance matrix of the system;

The calculation formula of H _θ,k is as follows

in

in, They are the prior prediction vectors of the parameters of this cycle/> The first, second, third, and fourth elements in . Similarly, /> are the posterior estimation vectors of the state in the previous cycle respectively/> First, second element.

It should be noted, K _x,k-1 and/> It needs to be obtained through the iteration of the previous step. When k=0, the initial values of these three quantities can all be taken as zero matrices.

State posterior estimate:

in, represents the posterior estimation covariance matrix of the state of the current cycle, K _x,k represents the Kalman gain of the state of the current cycle, and H _x,k represents the value of the derivative of the discrete output equation with respect to the motor state X* in the current cycle.

In the above formulas (13) to (18), the values of the system noise covariance matrix _Qp in the parameter prior prediction and the system noise covariance matrix _Qx in the state prior prediction mainly affect the convergence speed of the system parameter observation. Optionally, in this embodiment, the value of _Qp is 0.1 times the parameter estimate, that is:

Q _P = 0.1·diag([R _s T _dd T _dq T _qq ]) (19)

M mainly depends on the noise of current sampling in the actual electric drive system.

The full-parameter observer based on Kalman filtering designed in this embodiment, in which the discretized motor state model f ^* () is used, fully considers the high-coupling and strong nonlinear characteristics of the permanent magnet synchronous motor, as well as the influence of the injected signal on the motor state when a high-frequency disturbance voltage signal is injected. Therefore, the motor state model f ^* () can more accurately reflect the state of the motor during operation, is not limited by the motor speed, and can achieve accurate observation of the motor parameters under all working conditions. In addition, in this embodiment, two Kalman filters are designed for observing the motor state and motor parameters respectively, and the prior prediction results of the two Kalman filters will be called by each other. Under the mutual promotion of state observation and parameter observation, the accuracy of parameter observation can be further improved, thereby improving the accuracy of subsequent motor parameter identification.

Based on the full parameter observer based on Kalman filtering provided in the above embodiment, in another embodiment of the present invention, a method for identifying full parameters of a permanent magnet synchronous motor is provided, and the identified parameters include: motor stator winding resistance R _s , d-axis incremental self-inductance Q-axis incremental self-inductance/> And the incremental mutual inductance of d and q axes/> The d-axis magnetic flux is ψ _d0 and the q-axis magnetic flux is ψ _q0 .

As shown in FIG. 2 and FIG. 3 , this embodiment includes:

Small signal disturbance voltage injection step: during the operation of the motor, a d-axis small signal disturbance voltage u _dh and a q-axis small signal disturbance voltage u _qh with a frequency of ω _h are respectively injected into the d-axis and q-axis command voltages output by the current loop of the motor; in order to avoid generating odd harmonics in the ABC three-phase of the motor, in this embodiment, ω _h ≠ 6nω _e , ω _e represents the motor fundamental frequency, and n is a positive integer ω _e represents the motor fundamental frequency;

The full parameter identification steps include:

(S1) sampling the three-phase current of the motor in the running state, and transforming it to a dq-axis rotating coordinate system whose synchronous speed is the motor fundamental frequency ω _e , to obtain the d-axis current i _d and the q-axis current i _q ;

(S2) extracting the components with frequency _ωh from the _d -axis current id and the q-axis current _iq respectively, and obtaining the d-axis small-signal disturbance current _idh and the q-axis small-signal disturbance current _iqh under the excitation of the d-axis small-signal disturbance voltage _udh and the q-axis small-signal disturbance voltage _uqh ;

Specifically, the d-axis small-signal disturbance current i _dh and the q-axis small-signal disturbance current i _qh can be extracted through a bandpass filter with a frequency of ω _h ;

(S3) Input the d-axis small signal disturbance current i _dh and the q-axis small signal disturbance current i _qh into the full parameter observer based on Kalman filtering, and obtain the parameter posterior estimation value θ = [R _s T _dd T _dq T _qq ] ^T output by the full parameter observer based on Kalman filtering to calculate the motor stator winding resistance R _s and the d-axis incremental self-inductance Q-axis incremental self-inductance/> And the incremental mutual inductance of d and q axes

Since the observer provided in the above embodiment can accurately observe and obtain the a posteriori estimated value θ of the motor parameter under all operating conditions, the present embodiment can accurately identify the resistance parameter and inductance parameter of the motor under all operating conditions;

(S4) Based on the parameters calculated in (S3), calculate the d-axis magnetic flux ψ _d0 and the q-axis magnetic flux ψ _q0 ;

In order to ensure accurate identification of the d-axis flux ψ _d0 and the q-axis flux ψ _q0 , in this embodiment, after obtaining the resistance parameters and the inductance parameters based on the parameter identification of the observer output, these parameters are substituted into the above formula (6), and the calculation formulas of the d-axis flux ψ _d0 and the q-axis flux ψ _q0 can be obtained as follows:

Wherein, u _d0 and u _q0 represent the d-axis fundamental frequency signal voltage and the q-axis fundamental frequency signal voltage, respectively, i _d0 and i _q0 represent the d-axis fundamental frequency signal current and the q-axis fundamental frequency signal current, respectively;

Since the above formula (6) fully considers the influence of the injected small signal disturbance voltage, the present embodiment can accurately identify the d-axis flux ψ _d0 and the q-axis flux ψ _q0 .

In this embodiment, between steps (S1) and (S2), the following steps are also included:

Filter out the components with a frequency of ω _h in the d-axis current i _d and the q-axis current i _q to obtain the d-axis DC current i _d0 and the q-axis DC current i _q0 in the motor running state; specifically, a band-stop filter with a frequency of ω _h can be used to filter out the components with a frequency of ω _h in the d-axis current i _d and the q-axis current i _q ;

Input the d-axis DC current i _d0 and the q-axis DC current i _q0 into the current loop;

After injecting a small signal disturbance voltage into the current loop in this embodiment, the d-axis current i _d and the q-axis current i _q are obtained by sampling and coordinate transformation of the three-phase current of the motor, which include two parts: the DC currents i _d0 and i _q0 that generate torque when the motor is running, and the small signal currents i _dh and i _qh with a frequency of ω _h . Before the d- and q-axis currents are input into the current loop in this embodiment, the components with a frequency of ω _h in the d- and q-axis currents are first filtered out by filtering, which can prevent the current loop output command voltage from generating a suppression voltage signal that differs by 180° from the rotating voltage signal, thereby avoiding the impact of the injected rotating high-frequency voltage on the motor control part, ensuring the normal operation of the motor, and at the same time avoiding the suppression signal generated by the current loop from interfering with the motor parameter identification, thereby ensuring the accuracy of the motor parameter identification.

It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims (4)

1. The full-parameter observer based on Kalman filtering is used for carrying out parameter identification on a permanent magnet synchronous motor, and small signal voltages u _dh and u _qh with the frequency of omega _h are respectively injected into d and q-axis command voltages output by a current loop of the permanent magnet synchronous motor in the operation process; wherein the full parameter observer comprises: a parameter extended kalman filter and a state extended kalman filter;

The parameter extended kalman filter includes: the parameter prior prediction module and the parameter posterior estimation module; the state-extended kalman filter includes: the state prior prediction module and the state posterior estimation module;

the first input end of the parameter priori prediction module is connected to the output end of the parameter posterior estimation module, and is used for performing priori prediction on the parameters of the current period to obtain the parameter priori predicted value of the current period

The state prior prediction module has a first input end connected to the output end of the state posterior estimation and a second input end connected to the output end of the parameter prior prediction module, and is used for carrying out prior prediction on the state of the current period to obtain a state prior prediction value of the current period

The first input end of the parameter posterior estimation module is connected to the output end of the parameter prior prediction module, the second input end of the parameter posterior estimation module is connected to the output end of the state prior prediction module, and the parameter posterior estimation module is used for carrying out posterior estimation on the parameters of the current period to obtain the parameter posterior estimation value of the current period

The state posterior estimation module has a first input end connected to the output end of the state prior prediction module and a second input end connected to the output end of the parameter prior prediction module, and is used for performing posterior estimation on the state of the current period to obtain a state posterior estimation value of the current period

Wherein,The state posterior estimate value of the previous cycle is represented, and U _k-1 represents the small signal voltage injected in the previous cycle; t _s denotes the sampling time interval, ω _e denotes the motor fundamental frequency; /(I)R _s represents the resistance of the motor stator winding; Expressed as an inverse of the inductance matrix,/> AndRespectively represent d-axis increment self-inductance and q-axis increment self-inductance,/>The d and q axes incremental mutual inductance is represented; θ= [ R _s T_dd T_dq T_qq]^T ] represents the parameter to be observed,/>F ^* () represents the discrete state equation of the motor;

The calculation expression of the parameter prior prediction module for signal processing is as follows:

Wherein, A parameter posterior estimate representing a previous period; /(I)Parameter prior prediction covariance matrix representing current period,/>The parameter posterior estimation covariance matrix of the last period is represented, and Q _p represents the system noise covariance matrix in the parameter prior prediction;

The state prior prediction model carries out signal processing and has the following calculation expression:

Wherein, Parameter prior estimation covariance matrix representing current period,/>The state posterior estimation covariance matrix of the last period is represented, and Q _x represents the system noise covariance matrix in state prior prediction; f _x,k-1 represents the value of the derivative of the discrete state equation F ^* () on the motor state X in the last cycle;

The calculation expression of the signal processing by the parameter posterior estimation module is as follows:

Wherein, Representing the output of the motor in the current period, wherein h ^* () represents a discrete output equation of the motor; /(I)A parameter posterior estimation covariance matrix representing a current period; k _θ,k represents the Kalman gain of the current period parameter, H _θ,k represents the value of the derivative of the discrete output equation on the motor parameter theta in the current period; m represents a measurement noise covariance matrix of the system;

The state posterior estimation module performs signal processing according to the following calculation expression:

Wherein, The state posterior estimation covariance matrix of the current period is represented, K _x,k represents the kalman gain of the state of the current period, and H _x,k represents the value of the derivative of the discrete output equation on the state X of the motor in the current period.

2. A method for identifying all parameters of a permanent magnet synchronous motor based on the kalman filter-based all-parameter observer as claimed in claim 1, comprising:

a small signal disturbance voltage injection step: in the running process of the motor, d-axis small signal disturbance voltage u _dh and q-axis small signal disturbance voltage u _qh with the frequency of omega _h are respectively injected into d-axis command voltage and q-axis command voltage output by a current loop of the motor;

The full parameter identification step comprises the following steps:

(S1) sampling three-phase current in a motor running state, and transforming the three-phase current into a dq-axis rotation coordinate system with synchronous rotation speed of the fundamental frequency omega _e of the motor to obtain d-axis current i _d and q-axis current i _q;

(S2) extracting components with frequency ω _h from the d-axis current i _d and the q-axis current i _q, respectively, to obtain a d-axis small signal disturbance current i _dh and a q-axis small signal disturbance current i _qh under excitation of the d-axis small signal disturbance voltage u _dh and the q-axis small signal disturbance voltage u _qh;

(S3) inputting the d-axis small signal disturbance current i _dh and the q-axis small signal disturbance current i _qh into the Kalman filtering-based full-parameter observer to obtain a parameter posterior estimation value theta= [ R _s T_dd T_dq T_qq]^T output by the Kalman filtering-based full-parameter observer so as to calculate the motor stator winding resistance R _s and d-axis increment self-inductance Q-axis delta self-inductance/>Incremental mutual inductance of d and q axes/>

(S4) calculating d-axis flux linkage ψ _d0 and q-axis flux linkage ψ _q0 based on the parameters calculated in (S3).

3. The method of claim 2, wherein in the step (S4), the d-axis flux linkage ψ _d0 and the q-axis flux linkage ψ _q0 are calculated as:

Where u _d0 and u _q0 represent a d-axis fundamental frequency signal voltage and a q-axis fundamental frequency signal voltage, respectively, and i _d0 and i _q0 represent a d-axis fundamental frequency signal current and a q-axis fundamental frequency signal current, respectively.

4. A method of full parameter identification for a permanent magnet synchronous motor according to claim 2 or 3, further comprising, between said steps (S1) and (S2):

Filtering out components with the frequency omega _h in the d-axis current i _d and the q-axis current i _q to obtain d-axis direct current i _d0 and q-axis direct current i _q0 in the motor running state;

The d-axis direct current i _d0 and the q-axis direct current i _q0 are input to the current loop.