CN106602952A - Flux linkage full-rank identification method for permanent magnet of PMSM - Google Patents

Abstract

The invention discloses a flux linkage full-rank identification method for a permanent magnet of a PMSM. According to the method, an unscented-Kalman-filter-algorithm-based parameter sensitivity analysis for the flux linkage identification precision of a permanent magnet of a PMSM is carried out; a permanent magnet flux linkage full-rank identification method of different rotating speed zones of a PMSM drive system is determined. With the unscented-Kalman-filter-algorithm-based permanent-magnet flux linkage full-rank identification method of different rotating speed zones of a PMSM drive system, influences on the permanent magnet flux linkage identification precision by the system measurement noises, PMSM parameter changing, and a degenerate-rank phenomenon of an identification model can be eliminated; linear error occurrence and complicated Jacobian matrix calculation during the state identification process of the traditional extended Kalman filter algorithm can be avoided; and the algorithm implementation difficulty can be reduce. Meanwhile, the number of to-be-identified parameters can be reduced at a high-speed operating zone in a PMSM drive system; the operand of the identification algorithm can be reduced; and reasonable consideration of the identification precision and the identification speed can be realized.

Description

A kind of PMSM permanent magnet flux linkages full rank discrimination method

Technical field

The present invention relates to the technical field of permagnetic synchronous motor (PMSM), and in particular to a kind of PMSM permanent magnet flux linkages full rank Discrimination method.

Background technology

The technical advantage such as PMSM has simple structure, fault rate low and operational efficiency is high, gradually in industrial servo-drive, new The fields such as energy automobile are applied widely.However, for many applications such as electric automobile, PMSM power densities Height, radiating condition is poor, and operating condition is complicated, under jointly controlling in torque capacity electric current ratio and weak magnetic more, there is stronger electricity Pivot reacts, in addition the factor such as natural aging, permanent magnet uniformly demagnetization or local demagnetization failure easily occurs, causes motor output to turn Square is reduced and torque pulsation, directly affects the direct torque precision and operational reliability of power drive system.

Meanwhile, in the control algolithm such as MTPA controls, weak magnetic control in PMSM, it is both needed to accurate permanent magnet magnetic chain information. Therefore, in order to realize PMSM drive system permanent magnets health status monitor and high performance controller design, it is necessary to it is accurate to obtain Permanent magnet flux linkage.

In order to accurately obtain PMSM permanent magnet flux linkages, French scholar Henwood N and Shandong University Wang Song are respectively adopted dragon Burger observer realizes the observation of PMSM permanent magnet flux linkages with method of least square, yet with observed result to the quick of measurement noise Perception, limits application of the program in actual industrial system.In order to solve the problems, such as noise jamming, Tsing-Hua University Xiao Xi professors It is assumed that on the premise of other PMSM parameter constants, being estimated online to permanent magnet flux linkage using expanded Kalman filtration algorithm Meter, but affected by magnetic circuit saturation and operation temperature rise, stator resistance R_sAnd d-q axle stator inductance L_dAnd L_qCan drive with PMSM There is different degrees of change in the change of dynamic system conditions, affects the estimated accuracy of permanent magnet flux linkage.

Need to be state variable by the PMSM parameter processings of change to ensure PMSM permanent magnet flux linkage identification precisions, realize Permanent magnet flux linkage identification under Parameters variation constraint.For this purpose, peace group's great waves is established recognize PMSM stator resistance R simultaneously_s, d-q axles Stator inductance L_dAnd L_qAnd permanent magnet flux linkage ψ_fAdaptive model；The electricity then based on PMSM q axles such as Cortes-Romero J A Pressure equation, all electromagnetic parameters of PMSM are realized using algebraically identification algorithm in the case of participating in without the need for parameter of electric machine initial value On-line identification.However, the order for realizing the PMSM state equations that permanent magnet flux linkage is recognized is 2, while two parameters can only be realized Full rank is recognized, stator resistance R_s, stator inductance L_dAnd L_qAnd permanent magnet flux linkage ψ_fWhile identification exist identification equation owe order ask Topic, the uniqueness of identification result lacks theory support, easily causes identification result to be absorbed in local optimum and even dissipates.For this purpose, northern Capital Aero-Space university professor Wang Lina proposes a kind of based on model reference adaptive calculation for face mounted permagnetic synchronous motor The step identification method of method, the method is first with d shaft voltage equation estimation PMSM armature inductance L_s, recycle the L for obtaining_sIt is real Existing ψ_fWith R_sSynchronous full rank identification.Due to adopting i face mounted permagnetic synchronous motor more_d=0 control, in order to realize ψ_f、R_sIt is same Step full rank identification, need to inject amplitude and the rational d axles current perturbation of frequency, and not account in PMSM drive system runnings L_sChange is to ψ_fThe impact of identification precision.For this purpose, France scholar Underwood S J are based on IPM synchronous motor R_s、 L_d、L_q、ψ_fThe different time scales of four parameters, are divided into fast varying parameter and slow varying parameter, and using two different time chis The least-squares algorithm of degree realizes the on-line identification of aforementioned four parameter, although the deficient order that the method can solve identification model is asked Topic, but in order to ensure algorithmic statement, still need to inject amplitude and the rational current perturbation of frequency to guarantee slow time scale most in d axles The identification precision of young waiter in a wineshop or an inn's multiplication algorithm, and identification result is easily by measurement influence of noise.Artificial intelligence with evolution algorithm as representative by In with stronger Nonlinear Processing ability, there is certain application in the identification of PMSM permanent magnet flux linkages, but how to reduce it Full rank recognizes amount of calculation, is but still key technical problem urgently to be resolved hurrily.

In consideration of it, the PMSM permanent magnet flux linkages that systematic survey noise, PMSM Parameters variations, identification equation are owed under order constraint are high Precision, online, full rank identification have become the monitoring of PMSM permanent magnets health status and high performance control field pass urgently to be resolved hurrily Key problem.

The content of the invention

In order to solve above-mentioned technical problem, the present invention provides a kind of PMSM permanent magnet flux linkages full rank discrimination method, based on double Unscented kalman filtering algorithm carries out subregion joint full rank identification to PMSM permanent magnet flux linkages, realizes measurement noise, PMSM parameters PMSM permanent magnet flux linkages high accuracy, online, full rank identification under order constraint are owed in change, identification model, are to realize PMSM drivetrains The high performance control of system and the on-line monitoring of permanent magnet health status provide foundation.

In order to achieve the above object, the technical scheme is that：A kind of PMSM permanent magnet flux linkages full rank discrimination method, adopts Stator voltage u in the d-q shaftings of collection PMSM_d、u_q, stator current i_d、i_qAnd PMSM drive system rotor electrical angular velocity omegas_e, Setting up in d-q shaftings is used for the PMSM state equations of permanent magnet flux linkage identification, realizes based on Unscented kalman filtering algorithm The parameters sensitivity analysis of PMSM permanent magnet flux linkage identification precisions, based on analysis result PMSM drive system different rotating speeds area is determined The permanent magnet flux linkage full rank discrimination method based on double Unscented kalman filtering algorithms, realize measurement noise, permagnetic synchronous motor Parameters variation and identification model owe permanent magnet flux linkage high accuracy, online, full rank identification under order constraint；Its step is as follows：

Stator voltage u in step one, the d-q shaftings of collection PMSM_dAnd u_q, stator current i_dAnd i_qAnd PMSM drivetrains System rotor electrical angular velocity omega_e；

Step 2, the PMSM state equations set up in d-q shaftings, based on Unscented kalman filtering algorithm PMSM permanent magnetism is realized Body magnetic linkage is recognized, under analysis PMSM drive system difference operating conditions, stator resistance R_s, d-q axle stator inductance L_dAnd L_qChange is right The impact of permanent magnet flux linkage identification precision, realizes the parameter based on the permanent magnet flux linkage identification precision of Unscented kalman filtering algorithm Sensitivity analyses；

Step 3, according to step 2 result, in PMSM drive system low cruises area, by d-q axle stator inductance L_d、L_qConnection Close identification and stator resistance R_s, permanent magnet flux linkage ψ_fJoint identification is combined, updated each other, under identification equation full rank state, is disappeared Except stator resistance R_s, d-q axle stator inductance L_dAnd L_qImpact of the change to permanent magnet flux linkage estimated accuracy；In PMSM drive systems Middle and high fast Operational Zone, using permanent magnet flux linkage ψ_fIdentification and d-q axle stator inductance L_d、L_qJoint identification combines, in identification side Under journey full rank state, d-q axle stator inductance L are eliminated_d、L_qImpact of the change to permanent magnet flux linkage identification precision.

Stator voltage u in the d-q shaftings of the PMSM_dAnd u_q, stator current i_dAnd i_qAcquisition methods be：Sampling PMSM Stator line voltage u_ab、u_bc, three-phase current i_a、i_b、i_c, and obtained by coordinate transform, transformation matrix of coordinates is respectively：

In formula, θ is flux linkage position of the rotor angle.

Stator voltage u in the d-q shaftings of the PMSM_dAnd u_q, stator current i_dAnd i_qAcquisition methods be：Directly adopt The d-q shaft voltage command values that PMSM driving system controllers are calculatedWithReplace d-q axle stator voltages u_dAnd u_q, d-q axles Current instruction valueReplace d-q axle stator current i_dAnd i_q。

The acquisition methods of the PMSM drive systems rotor electrical angular velocity are obtained by incremental optical-electricity encoder, step Suddenly it is：

(1) in t₁And t₂Umber of pulse N that optical rotary encoder sends on the d-q axles of neighbouring sample instance sample PMSM₁、 N₂, sampling instant t₁And t₂Difference be sampling period T；

(2) according to rotor angular rate ω_eWith optical rotary encoder impulse sampling value N₁、N₂And between sampling period T Relation calculate rotor angular rate ω_e, its expression formula is：

In formula, M is the optical rotary encoder umber of pulse of a week, and p is permagnetic synchronous motor number of pole-pairs.

It is used to realize that the method for PMSM state equations that permanent magnet flux linkage is recognized is in the d-q shaftings：

In formula, u_dAnd u_qD-q axle stator voltages, i are represented respectively_dAnd i_qD-q axle stator currents, L are represented respectively_dAnd L_qRespectively Represent d-q axle stator inductances；R_sRepresent stator resistance, ψ_fRepresent permanent magnet flux linkage, ω_eRepresent rotor electric angle frequency；

In PMSM state equations, state vector is：X=[i_d i_q ψ_f]^T, input quantity is：U=[u_d/L_d u_q/L_q]^T, output Measure and be：Y=[i_d i_q]^T。

It is described to realize that the method that PMSM permanent magnet flux linkages are recognized is based on Unscented kalman filtering algorithm：Permanent magnet will be realized The PMSM state equations of magnetic linkage identification are described as the measurement equation table of the general type of nonlinear system, state equation and discretization It is shown as：

In formula, x (t) be system mode vector, y (t_k) it is output, f () represents systematic state transfer equation, h () Represent system measuring equation, σ (t) is the process noise for considering model uncertainty and measuring uncertainty, μ (t_k) it is to consider mould Uncertain and measuring uncertainty the measurement noise of type, the variance matrix of σ (t) is Q (t), μ (t_k) variance matrix be R T (), u (t) is definitiveness input vector, B represents control matrix；

Above-mentioned PMSM state equations, state vector, input vector, output vector are substituted into into Unscented kalman filtering algorithm In, realize based on the state vector recursion of Unscented kalman filtering algorithm, you can realize the on-line identification of PMSM permanent magnet flux linkages, Step is：

(1) state vector initialization

Process noise covariance matrix Q and measurement noise covariance matrix R initial values are set according to priori, shape is initialized State vectorInit state varivance matrix

(2) Sigma points are calculated

Pass through within each sampling periodCalculate Sigma Point (k=1,2, ∞) and, obtain the Sigma dot matrixs of n × (2n+1)；Wherein：χ_k-1For Sigma dot matrixs,P_k-1It is expressed as the prediction average and covariance of a moment state vector, n is state vector dimension, λ=α²(n+ B) it is scale parameter, α is dispersion level of the Sigma points near state variable average, and b is scale coefficient；

(3) time renewal

Realize that Sigma points are transmitted by the system state equation of discretization：I.e.According to Transmission result obtains the prediction average and covariance of state vector：

Wherein,The average weight of state vector is represented,The covariance weight of state vector is represented,WithQuantitative relation be：

(4) measurement updaue

According to system measurement data, by formula It is capable of achieving the one-step prediction of state vector Update and updated with Kalman gain, obtain the optimal estimation of state vector and variance matrix；Wherein, H represents calculation matrix, K_kTable Show Kalman gain；

K=k+1 is made, the iteration output of state vector is realized in repeat step (2)-(4).

It is described to realize based on the parameters sensitivity analysis of the PMSM permanent magnet flux linkage identification precisions of Unscented kalman algorithm Method is：It is determined that the PMSM stator resistance R affected by system conditions and operation temperature rise_s, d axle inductances L_d, q axle inductances L_q's Excursion, and the permanent magnet flux linkage identification under different system operating mode based on Unscented kalman filtering algorithm is analyzed within this range Error.

In the PMSM drive systems low cruise region, full rank identification equation is：

In the middle and high fast operation area of the PMSM drive systems, full rank identification equation is：

Beneficial effect of the present invention：Compared with prior art, the present invention can owe under order constraint, to realize PMSM in identification model The full rank identification of permanent magnet flux linkage, eliminates measurement noise, PMSM Parameters variations and identification model and owes order to permanent magnet flux linkage identification The impact of precision；Avoid simultaneously linearized stability of the conventional Extension Kalman filtering algorithm in state vector identification process and The calculating of complicated Jacobian matrix, reduces algorithm and realizes difficulty while identification precision is improved；In PMSM drive systems, High-speed cruising area, can reduce parameter R to be identified on the premise of permanent magnet flux linkage identification precision is ensured_s, reduce identification and calculate Method amount of calculation.

Description of the drawings

In order to be illustrated more clearly that the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing The accompanying drawing to be used needed for having technology description is briefly described, it should be apparent that, drawings in the following description are only this Some embodiments of invention, for those of ordinary skill in the art, on the premise of not paying creative work, can be with Other accompanying drawings are obtained according to these accompanying drawings.

Fig. 1 is the parameter sensitivity of the PMSM permanent magnet flux linkage identification precisions based on Unscented kalman filtering algorithm of the present invention Property analysis result.

Fig. 2 is the PMSM permanent magnet flux linkages subregion joint full rank identification based on double Unscented kalman filtering algorithms of the present invention The structured flowchart of method.

Fig. 3 is the permanent magnet flux linkage identification result of the middle and high fast Operational Zones of PMSM of the present invention.

Specific embodiment

Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete Site preparation is described, it is clear that described embodiment is only a part of embodiment of the invention, rather than the embodiment of whole.It is based on Embodiment in the present invention, it is every other that those of ordinary skill in the art are obtained under the premise of creative work is not paid Embodiment, belongs to the scope of protection of the invention.

A kind of PMSM permanent magnet flux linkages full rank discrimination method, step is as follows：

Step one, stator voltage u obtained in the d-q shaftings of PMSM_dAnd u_q, stator current i_dAnd i_qAnd PMSM drivetrains System rotor electrical angular velocity omega_e。

Stator voltage u in the d-q shaftings of the PMSM_dAnd u_q, stator current i_dAnd i_qAcquisition methods have following two：

(1) stator line voltage u of sampling PMSM_ab、u_bc, three-phase current i_a、i_b、i_c, and obtained by coordinate transform.Coordinate Transformation matrix is respectively：

In formula, θ is flux linkage position of the rotor angle.

(2) the d-q shaft voltage command values directly calculated using PMSM driving system controllersWithReplace d-q axles Stator voltage u_dAnd u_q, d-q shaft current command valuesReplace d-q axle stator current i_dAnd i_q。

Due to adopting electric current, rotating speed double circle structure PMSM drive systems more, therefore, method (1) needs to increase voltage adopts Sample and isolation circuit, increase system hardware expense；And method (2) using command value substitute actual value, though without the need for voltage sample with Isolation circuit, but need to consider that inverter is non-linear and sample circuit time lag caused by substitute offset issue, needed if necessary Effect compensation.

The acquisition methods of the PMSM drive systems rotor electrical angular velocity are obtained by incremental optical-electricity encoder, step Suddenly it is：

(1) in t₁And t₂Umber of pulse N that optical rotary encoder sends on the d-q axles of neighbouring sample instance sample PMSM₁、 N₂, sampling instant t₁And t₂Difference be sampling period T.

(2) according to rotor angular rate ω_eWith optical rotary encoder impulse sampling value N₁、N₂And between sampling period T Relation calculate rotor angular rate ω_e, its expression formula is：

In formula, M is the optical rotary encoder umber of pulse of a week, and p is permagnetic synchronous motor number of pole-pairs.

Step 2, the PMSM state equations set up in d-q shaftings, based on Unscented kalman filtering algorithm PMSM permanent magnetism is realized Body magnetic linkage is recognized, under analysis PMSM drive system difference operating conditions, stator resistance R_s, stator inductance L_d、L_qChange is to permanent magnet Magnetic linkage ψ_fThe impact of identification precision, realizes based on the permanent magnet flux linkage ψ of Unscented kalman filtering algorithm_fThe parameter of identification precision is quick Perceptual analysis；

It is used to realize that the PMSM state equations that permanent magnet flux linkage is recognized are in the d-q shaftings：

Wherein, u_dAnd u_qD-q axle stator voltages, i are represented respectively_dAnd i_qD-q axle stator currents, L are represented respectively_dAnd L_qRespectively Represent d-q axle stator inductances；R_sRepresent stator resistance, ψ_fRepresent permanent magnet flux linkage, ω_eRepresent rotor electric angle frequency.

For the state equation of formula (3) description, its state vector x, input quantity u and output y are respectively：

X=[i_d i_q ψ_f]^T, u=[u_d/L_d u_q/L_q]^T, y=[i_d i_q]^T (4)

It is described to realize that the method that PMSM permanent magnet flux linkages are recognized is based on Unscented kalman filtering algorithm：

The PMSM state equations for realizing permanent magnet flux linkage identification that formula (3) is described are characterized as into the one of nonlinear system As the measurement equation of form, its state equation and discretization be expressed as

In formula：X (t) be system state variables, y (t_k) it is output, f () represents systematic state transfer equation, h () Represent system measuring equation, σ (t), μ (t_k) respectively consider the process noise of model uncertainty and measuring uncertainty and survey Amount noise, the variance matrix of σ (t) is Q (t), μ (t_k) variance matrix be R (t), u (t) be definitiveness input vector, B for control Matrix processed, is constant value matrix.

PMSM state equations, state vector, input vector, output vector are substituted in Unscented kalman filtering algorithm, it is real Now it is based on the state vector recursion of Unscented kalman filtering algorithm, you can be capable of achieving PMSM permanent magnet flux linkage ψ_fOn-line identification, Step is：

(1) state vector initialization

Process noise covariance matrix Q and measurement noise covariance matrix R initial values are set according to priori, and is initialized State vector x and State error variance matrix P：

(2) Sigma points are calculated

Within each sampling period (k=1,2, ∞) and Sigma points are calculated according to formula (7), obtain a n × (2n + 1) Sigma dot matrixs.

In formula：χ_k-1For Sigma dot matrixs,P_k-1The prediction average and covariance at a moment are expressed as, n is State vector dimension, λ=α²(n+b) it is scale parameter；α is dispersion level of the Sigma points near state variable average, is determined Sigma point distribution situations, are usually taken to be interval [10^-4, 1] on little positive number；B is scale coefficient, is usually taken to be 0 or 3-n.

(3) time renewal

The transmission of Sigma points should be realized by the system state equation of discretization, as shown in formula (8), according to transmission result The prediction average and its covariance of state vector are obtained, as shown in formula (9).

Wherein,The average weight of state vector is represented,Represent the covariance weight of state vector, quantitative relation For：

(4) measurement updaue

According to measurement data, it is capable of achieving one-step prediction by formula (11)-(16) and Kalman gain updates, obtains state The optimal estimation of vector and its variance matrix.

(5) k=k+1 is made, the iteration output of state vector is realized in duplication stages (2)-(4).

Formula (5) is substituted in the Unscented kalman filtering algorithm flow described by formula (6)-(16), you can realize being based on nothing The state vector recursion of mark Kalman filtering algorithm.

It is described to realize based on the PMSM permanent magnet flux linkage ψ of Unscented kalman algorithm_fThe parameters sensitivity analysis of identification precision Method be：It is determined that the PMSM stator resistance R affected by operating condition and operation temperature rise_sAnd d-q axle stator inductance L_d、L_qChange Change scope, and analyze the ψ under different system operating mode based on Unscented kalman filtering algorithm within this range_fIdentification Errors.

Step 3, according to step 2 conclusion, in PMSM drive system low cruises area, by d-q axle stator inductance L_d、 L_qCarry out joint full rank identification, and with stator resistance R_s, permanent magnet flux linkage ψ_fThe identification of joint full rank combines, updates each other, is distinguishing Under knowing equation full rank state, stator resistance R is eliminated_s, d-q axle stator inductance L_d、L_qChange is to permanent magnet flux linkage ψ_fIdentification precision Affect, shown in full rank identification equation such as formula (16).

Further, in the middle and high fast Operational Zone of PMSM drive systems, due to stator resistance R_sChange is to permanent magnet flux linkage ψ_f The impact of identification precision is less, therefore using permanent magnet flux linkage ψ_fIdentification and d-q axle stator inductance L_d、L_qThe identification of joint full rank is mutually tied The method of conjunction, eliminates d-q axle stator inductance L_d、L_qChange is to permanent magnet flux linkage ψ_fThe impact of identification precision, full rank identification equation is such as Shown in formula (17).

According to above-mentioned thinking determine considerations PMSM Parameters variations when permanent magnet flux linkage subregion combine full rank discrimination method Structured flowchart is as shown in Figure 2.In PMSM drive system Operational Zone of the rotating speed less than 100rpm, using formula (16) Suo Shi, d-q axles are determined Sub- inductance L_d、L_qJoint full rank is recognized and stator resistance R_s, permanent magnet flux linkage ψ_fThe full rank identification side that the identification of joint full rank combines Method, eliminates R_s、L_d、L_qChange is to permanent magnet flux linkage ψ_fThe impact of identification precision, under the constraint of PMSM Parameters variations permanent magnet is realized Magnetic linkage ψ_fFull rank identification.In PMSM drive system middle and high fast Operational Zone of the rotating speed more than 100rpm, using formula (17) Suo Shi, D-q stator inductance L_d、L_qJoint full rank is recognized and permanent magnet flux linkage ψ_fThe full rank discrimination method that identification combines, eliminates d-q stators Inductance L_d、L_qChange is to permanent magnet flux linkage ψ_fThe impact of estimated accuracy, under the constraint of PMSM Parameters variations permanent magnet flux linkage ψ is realized_f Full rank identification.

Carried out experimental verification to the method for the present invention, experiment condition is given load torque 3Nm, and take rotating speed from 900 revs/min of dynamic processes for being down to 450 revs/min.In view of under laboratory environment, parameter of electric machine change is less in the short time, this The bright checking that aforementioned high speed scheme is completed near motor parameters value.Stator resistance R_sWhen being taken as the 150% of design load, D axle stator inductance L_d, q axle stator inductance L_qAnd permanent magnet flux linkage ψ_fIdentification result is respectively such as Fig. 3 (a), 3 (b) and Fig. 3 (c) institutes Show.With PMSM design load (L_d=L_q=0.001283H, ψ_f=0.1278Wb) compare, the identification precision of presently disclosed method Higher, Identification Errors average can be controlled within 7%, and in the middle and high fast Operational Zone of PMSM drive systems, without the need for considering stator Resistance R_sImpact of the change to permanent magnet flux linkage identification precision, reduces parameter stator resistance R to be identified_s, effectively reduce identification Algorithm amount of calculation.

The above, the only present invention preferably specific embodiment, but protection scope of the present invention is not limited thereto, Any those familiar with the art the invention discloses technical scope in, the change or replacement that can be readily occurred in, All should be included within the scope of the present invention.

Claims (9)

1. a kind of PMSM permanent magnet flux linkages full rank discrimination method, it is characterised in that the stator voltage in the d-q shaftings of collection PMSM u_d、u_q, stator current i_d、i_qAnd PMSM drive system rotor electrical angular velocity omegas_e, setting up in d-q shaftings is used for permanent magnet flux linkage The PMSM state equations of identification, realize that the parameter based on the PMSM permanent magnet flux linkage identification precisions of Unscented kalman filtering algorithm is quick Perceptual analysis, based on analysis result determine PMSM drive system different rotating speeds area based on double Unscented kalman filtering algorithms forever Magnet magnetic linkage full rank discrimination method, realizes that measurement noise, permagnetic synchronous motor Parameters variation and identification model are owed under order constraint Permanent magnet flux linkage high accuracy, online, full rank identification；Its step is as follows：

Stator voltage u in step one, the d-q shaftings of collection PMSM_dAnd u_q, stator current i_dAnd i_qAnd PMSM drive system rotors Electric angular velocity omega_e；

Step 2, the PMSM state equations set up in d-q shaftings, based on Unscented kalman filtering algorithm PMSM permanent magnet magnetics are realized Chain is recognized, under analysis PMSM drive system difference operating conditions, stator resistance R_s, d-q axle stator inductance L_dAnd L_qChange is to permanent magnetism The impact of body magnetic linkage identification precision, realizes the parameter sensitivity based on the permanent magnet flux linkage identification precision of Unscented kalman filtering algorithm Property analysis；

Step 3, according to step 2 result, in PMSM drive system low cruises area, by d-q axle stator inductance L_d、L_qCombine and distinguish Know and stator resistance R_s, permanent magnet flux linkage ψ_fJoint identification is combined, updated each other, under identification equation full rank state, eliminates fixed Sub- resistance R_s, d-q axle stator inductance L_d、L_qImpact of the change to permanent magnet flux linkage estimated accuracy；It is middle and high in PMSM drive systems Fast Operational Zone, using permanent magnet flux linkage ψ_fIdentification and d-q axle stator inductance L_d、L_qJoint identification combines, in identification equation full rank Under state, d-q axle stator inductance L are eliminated_d、L_qImpact of the change to permanent magnet flux linkage identification precision.

2. PMSM permanent magnet flux linkages full rank discrimination method according to claim 1, it is characterised in that the d-q of the PMSM Stator voltage u in shafting_dAnd u_q, stator current i_dAnd i_qAcquisition methods be：Stator line voltage u of sampling PMSM_ab、u_bc, three-phase Electric current i_a、i_b、i_c, and obtained by coordinate transform, transformation matrix of coordinates is respectively：

$\left[\begin{array}{c}{u}_d\\ u_q\end{array}\right]=\left[\begin{array}{cc}cos\mathrm{\θ}& sin\mathrm{\θ}\\ -sin\mathrm{\θ}& cos\mathrm{\θ}\end{array}\right]\left[\begin{array}{cc}2/3& 1/3\\ 0& -\sqrt{3}/3\end{array}\right]\left[\begin{array}{c}{u}_{ab}\\ u_{bc}\end{array}\right],$

$\left[\begin{array}{c}{i}_d\\ i_q\end{array}\right]=\left[\begin{array}{cc}cos\mathrm{\θ}& sin\mathrm{\θ}\\ -sin\mathrm{\θ}& cos\mathrm{\θ}\end{array}\right]\left[\begin{array}{ccc}2/3& -1/3& -1/3\\ 0& -\sqrt{3}/3& \sqrt{3}/3\end{array}\right]\left[\begin{array}{c}{i}_a\\ i_b\\ i_c\end{array}\right];$

In formula, θ is flux linkage position of the rotor angle.

3. PMSM permanent magnet flux linkages full rank discrimination method according to claim 1, it is characterised in that the d-q of the PMSM Stator voltage u in shafting_dAnd u_q, stator current i_dAnd i_qAcquisition methods be：Directly calculated using PMSM driving system controllers D-q shaft voltage command values outWithReplace d-q axle stator voltages u_dAnd u_q, d-q shaft current command valuesReplace d- Q axle stator current i_dAnd i_q。

4. PMSM permanent magnet flux linkages full rank discrimination method according to claim 1, it is characterised in that the PMSM drivetrains The acquisition methods of system rotor electrical angular velocity are obtained by incremental optical-electricity encoder, and step is：

(1) in t₁And t₂Umber of pulse N that optical rotary encoder sends on the d-q axles of neighbouring sample instance sample PMSM₁、N₂, adopt Sample moment t₁And t₂Difference be sampling period T；

(2) according to rotor angular rate ω_eWith optical rotary encoder impulse sampling value N₁、N₂And the pass between sampling period T System calculates rotor angular rate ω_e, its expression formula is：

${\mathrm{\ω}}_e=\frac{N_2-N_1}{M\×T}\×p\×2\mathrm{\π};$

In formula, M is the optical rotary encoder umber of pulse of a week, and p is permagnetic synchronous motor number of pole-pairs.

5. PMSM permanent magnet flux linkages full rank discrimination method according to claim 1, it is characterised in that in the d-q shaftings For realizing that the method for PMSM state equations that permanent magnet flux linkage is recognized is：

$\left\{\begin{array}{c}\frac{{\mathrm{di}}_d}{dt}=\frac{u_d}{L_d}-\frac{R_s}{L_d}{i}_d+\frac{{\mathrm{\ω}}_e{L}_q}{L_d}{i}_q\\ \frac{{\mathrm{di}}_q}{dt}=\frac{u_q}{L_q}-\frac{R_s}{L_q}{i}_q-\frac{{\mathrm{\ω}}_e{L}_d}{L_q}{i}_d-\frac{{\mathrm{\ω}}_e{\mathrm{\ψ}}_f}{L_q}\\ \frac{{\mathrm{d\ψ}}_f}{dt}=0\end{array}\right.;$

In formula, u_dAnd u_qD-q axle stator voltages, i are represented respectively_dAnd i_qD-q axle stator currents, L are represented respectively_dAnd L_qRepresent respectively D-q axle stator inductances；R_sRepresent stator resistance, ψ_fRepresent permanent magnet flux linkage, ω_eRepresent rotor electric angle frequency；

In PMSM state equations, state vector is：X=[i_d i_q ψ_f]^T, input quantity is：U=[u_d/L_d u_q/L_q]^T, output For：Y=[i_d i_q]^T。

6. PMSM permanent magnet flux linkages full rank discrimination method according to claim 1, it is characterised in that described based on without mark card Kalman Filtering algorithm realizes that the method that PMSM permanent magnet flux linkages are recognized is：The PMSM state equations that permanent magnet flux linkage is recognized will be realized The measurement equation for being described as the general type of nonlinear system, state equation and discretization is expressed as：

$\left\{\begin{array}{c}\stackrel{\·}{x}\left(t\right)=f\[x\left(t\right)\]+Bu\left(t\right)+\mathrm{\σ}\left(t\right)\\ y\left(t_k\right)=h\[x\left(t_k\right)\]+\mathrm{\μ}\left(t_k\right)\end{array}\right.;$

In formula, x (t) be system mode vector, y (t_k) it is output, f () represents systematic state transfer equation, and h () is represented System measuring equation, σ (t) is the process noise for considering model uncertainty and measuring uncertainty, μ (t_k) it is to consider model not The measurement noise of definitiveness and measuring uncertainty, the variance matrix of σ (t) is Q (t), μ (t_k) variance matrix be R (t), u T () is definitiveness input vector, B represents control matrix；

Above-mentioned PMSM state equations, state vector, input vector, output vector are substituted in Unscented kalman filtering algorithm, it is real Now it is based on the state vector recursion of Unscented kalman filtering algorithm, you can realize the on-line identification of PMSM permanent magnet flux linkages, step For：

(1) state vector initialization

Process noise covariance matrix Q and measurement noise covariance matrix R initial values are set according to priori, init state to Amount x：Init state varivance matrix P：

(2) Sigma points are calculated

Pass through within each sampling periodCalculate Sigma Point (k=1,2, ∞) and, obtain the Sigma dot matrixs of n × (2n+1)；Wherein：χ_k-1For Sigma dot matrixs,P_k-1It is expressed as the prediction average and covariance of a moment state vector, n is state vector dimension, λ=α²(n+ B) it is scale parameter, α is dispersion level of the Sigma points near state variable average, and b is scale coefficient；

(3) time renewal

Realize that Sigma points are transmitted by the system state equation of discretization：I.e.According to transmission As a result the prediction average and covariance of state vector are obtained：

Wherein, W_i ^(c)Represent the average weight of state vector, W_i ^(m)Table Show the covariance weight of state vector, W_i ^(c)And W_i ^(m)Quantitative relation be：

$\left\{\begin{array}{c}{W}_0^{\left(m\right)}=\mathrm{\λ}/n+\mathrm{\λ}\\ W_0^{\left(c\right)}=\mathrm{\λ}/n+\mathrm{\λ}+(1-{\mathrm{\α}}^2+\mathrm{\β})\\ W_i^{\left(m\right)}=W_i^{\left(c\right)}=1/2(n+\mathrm{\λ})\end{array}\right.,i=1,2,...,2n;$

(4) measurement updaue

According to system measurement data, by formula It is capable of achieving the one-step prediction of state vector Update and updated with Kalman gain, obtain the optimal estimation of state vector and variance matrix；Wherein, H represents calculation matrix, K_kTable Show Kalman gain；

K=k+1 is made, the iteration output of state vector is realized in repeat step (2)-(4).

7. PMSM permanent magnet flux linkages full rank discrimination method according to claim 1, it is characterised in that the realization is based on nothing The method of the parameters sensitivity analysis of the PMSM permanent magnet flux linkage identification precisions of mark Kalman Algorithm is：It is determined that receiving system operation work The PMSM stator resistance R that condition and operation temperature rise affect_s, d axle inductances L_d, q axle inductances L_qExcursion, and divide within this range Permanent magnet flux linkage Identification Errors under analysis different system operating mode based on Unscented kalman filtering algorithm.

8. PMSM permanent magnet flux linkages full rank discrimination method according to claim 1, it is characterised in that the PMSM drivetrains In system low cruise region, full rank identification equation is：

$\left\{\begin{array}{c}\frac{{\mathrm{di}}_d}{dt}=\frac{u_d}{L_d}-\frac{R_s}{L_d}{i}_d+\frac{{\mathrm{\ω}}_e{L}_q}{L_q}{i}_q\\ \frac{{\mathrm{di}}_q}{dt}=\frac{u_q}{L_q}-\frac{R_s}{L_q}{i}_q-\frac{{\mathrm{\ω}}_e{L}_d}{L_q}{i}_d-\frac{{\mathrm{\ω}}_e{\mathrm{\ψ}}_f}{L_q}\\ \frac{{\mathrm{d\ψ}}_f}{dt}=0\\ \frac{{\mathrm{dR}}_s}{dt}=0\end{array}\right.,\left\{\begin{array}{c}\frac{{\mathrm{di}}_d}{dt}=\frac{u_d}{L_d}-\frac{R_s}{L_d}{i}_d+\frac{{\mathrm{\ω}}_e{L}_q}{L_q}{i}_q\\ \frac{{\mathrm{di}}_q}{dt}=\frac{u_q}{L_q}-\frac{R_s}{L_q}{i}_q-\frac{{\mathrm{\ω}}_e{L}_d}{L_q}{i}_d-\frac{{\mathrm{\ω}}_e{\mathrm{\ψ}}_f}{L_q}\\ \frac{{\mathrm{dL}}_d}{dt}=0\\ \frac{{\mathrm{dL}}_q}{dt}=0\end{array}\right..$

9. PMSM permanent magnet flux linkages full rank discrimination method according to claim 1, it is characterised in that the PMSM drivetrains Unite in middle and high fast operation area, full rank identification equation is：

$\left\{\begin{array}{c}\frac{{\mathrm{di}}_d}{dt}=\frac{u_d}{L_d}-\frac{R_s}{L_d}{i}_d+\frac{{\mathrm{\ω}}_e{L}_q}{L_q}{i}_q\\ \frac{{\mathrm{di}}_q}{dt}=\frac{u_q}{L_q}-\frac{R_s}{L_q}{i}_q-\frac{{\mathrm{\ω}}_e{L}_d}{L_q}{i}_d-\frac{{\mathrm{\ω}}_e{\mathrm{\ψ}}_f}{L_q}\\ \frac{{\mathrm{d\ψ}}_f}{dt}=0\end{array}\right.,\left\{\begin{array}{c}\frac{{\mathrm{di}}_d}{dt}=\frac{u_d}{L_d}-\frac{R_s}{L_d}{i}_d+\frac{{\mathrm{\ω}}_e{L}_q}{L_q}{i}_q\\ \frac{{\mathrm{di}}_q}{dt}=\frac{u_q}{L_q}-\frac{R_s}{L_q}{i}_q-\frac{{\mathrm{\ω}}_e{L}_d}{L_q}{i}_d-\frac{{\mathrm{\ω}}_e{\mathrm{\ψ}}_f}{L_q}\\ \frac{{\mathrm{dL}}_d}{dt}=0\\ \frac{{\mathrm{dL}}_q}{dt}=0\end{array}\right..$