CN109194225B - Online identification method for parameters of doubly-fed motor - Google Patents

Online identification method for parameters of doubly-fed motor Download PDF

Info

Publication number
CN109194225B
CN109194225B CN201811156781.8A CN201811156781A CN109194225B CN 109194225 B CN109194225 B CN 109194225B CN 201811156781 A CN201811156781 A CN 201811156781A CN 109194225 B CN109194225 B CN 109194225B
Authority
CN
China
Prior art keywords
formula
stator
innovation
doubly
rotor
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201811156781.8A
Other languages
Chinese (zh)
Other versions
CN109194225A (en
Inventor
吴定会
黄旭
纪志成
郑洋
黄海波
陶凯
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jiangnan University
Original Assignee
Jiangnan University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Jiangnan University filed Critical Jiangnan University
Priority to CN201811156781.8A priority Critical patent/CN109194225B/en
Publication of CN109194225A publication Critical patent/CN109194225A/en
Application granted granted Critical
Publication of CN109194225B publication Critical patent/CN109194225B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02P—CONTROL OR REGULATION OF ELECTRIC MOTORS, ELECTRIC GENERATORS OR DYNAMO-ELECTRIC CONVERTERS; CONTROLLING TRANSFORMERS, REACTORS OR CHOKE COILS
    • H02P21/00—Arrangements or methods for the control of electric machines by vector control, e.g. by control of field orientation
    • H02P21/14—Estimation or adaptation of machine parameters, e.g. flux, current or voltage

Abstract

The invention provides a method for identifying parameters of a doubly-fed motor on line, which aims at the characteristic of parameter change in the running process of the motor, is based on a multiple innovation identification theory and a random gradient identification method, and adds a variable forgetting factor, and provides a method for identifying parameters of the doubly-fed motor based on the variable forgetting factor multiple innovation random gradient method. The method considers the nonlinear strong coupling of the doubly-fed motor, adopts the stator flux linkage oriented vector control technology, builds the vector control system of the doubly-fed motor to collect data, and deducesdqAnd (3) in a standard form of the motor parameter identification model under the coordinate system, iteratively calculating the motor inductance and resistance parameters according to the method. The method is suitable for the double-fed motor in two states of electric and power generation, can identify the motor stator parameters on line, improves the control model precision and improves the control performance.

Description

Online identification method for parameters of doubly-fed motor
Technical Field
The invention relates to an online identification method for parameters of a double-fed motor, in particular to a double-fed motor suitable for a wind power generation system, and belongs to the technical field of motor control.
Background
The double-fed motor is a novel motor, and a speed regulating system with a larger capacity can be controlled by using a frequency converter with a smaller capacity in a specific occasion. The speed regulation device has a wider speed regulation range, can operate in various states, and can realize the regulation of the stator power factor. In addition, because the wind power generation has the characteristics of indefinite change of rotating speed, high randomness and the like, and the double-fed motor can realize variable-speed constant-frequency operation, the double-fed motor is widely applied to the field of wind power generation.
The vector control method is mature in a plurality of dual-feeder control methods, the vector control model is provided with two closed loops of a current loop and a speed loop, the motor is controlled to operate by setting parameters of two loop controllers, and the parameters of the controllers are set to be accurate motor parameters. Meanwhile, other control strategies such as direct torque control, adaptive control and the like do not leave accurate parameters, however, the control algorithm is inaccurate due to parameter uncertainty caused by changes of the doubly-fed motor and the surrounding environment and strong coupling among parameters in the model. In order to obtain accurate motor parameters to better control the operation of the motor, a suitable parameter identification algorithm needs to be found for parameter identification.
The relevant parameters of the doubly-fed motor are usually tested by an experimental method, and a no-load experiment and a locked rotor experiment are commonly used. The two methods have certain limitations, the motor parameters obtained by testing are only about different from real parameters, and some industrial equipment in industrial application has a plurality of limitations during operation, which is not beneficial to testing the motor parameters by an experimental method. The off-line parameter identification needs to be carried out before the motor runs, only the initial running parameters of the motor are obtained, and the change of the related parameters in the running of the motor cannot be obtained. If the motor parameters are acquired in real time and the control is adjusted according to the parameters, the change of the motor parameters needs to be changed in real time during the operation of the motor, so that the online parameter identification of the permanent magnet synchronous motor occurs. At present, the method for identifying the motor parameters mainly comprises a least square method, a model reference self-adaption method, a neural network method, a genetic algorithm and the like. The least square method is simple and high in precision, but the covariance matrix needs to be calculated in the identification process, so that the calculated amount is large; the design process of the self-adaptive law of the model reference self-adaptive method is complex, and particularly when a plurality of motors are identified simultaneously, the self-adaptive law meeting the stability requirement cannot be designed; the neural network method needs to stipulate a network structure in advance, and then determines a fitness function and the like through a genetic algorithm, so that the method is complex; the realization of the artificial intelligence method needs complex modeling processing, can increase the control complexity of the motor, and is not widely applied in actual engineering at present.
Therefore, in order to solve the problems that an off-line identification method is not easy to adopt when the doubly-fed motor actually operates, the existing on-line identification method has large calculation amount, complex design control and the like, and a doubly-fed motor parameter on-line identification method is needed to be designed. Aiming at the double-fed motor, a variable forgetting factor multi-innovation random gradient method is adopted to identify motor parameters in real time, so that the accuracy of a control model is improved, and the control performance is improved.
Disclosure of Invention
The invention aims to solve the problems that an off-line identification parameter identification method is not easy to adopt when a double-fed motor actually operates, the existing on-line identification method is large in calculated amount and complex in design and control processes, and the like, and covariance matrixes need to be calculated for identification by a least square method. The method is suitable for the double-fed motor in two states of electric operation and power generation, can identify the motor stator parameters on line, improves the control model precision and improves the control performance.
In order to achieve the purpose, the invention adopts the following technical scheme to realize the purpose:
a doubly-fed motor parameter online identification method comprises the following steps:
step one, establishing a mathematical model of the doubly-fed motor under a synchronous rotation dq coordinate system;
step two, improving the random gradient method, introducing an innovation length p, expanding the original single innovation quantity e (t) to a multi-innovation vector with the data length p, namely obtaining a multi-innovation random gradient algorithm according to the random gradient algorithm; adding a variable forgetting factor into the multi-innovation random gradient identification method to obtain a variable forgetting factor multi-innovation random gradient algorithm identification expression;
step three, transforming the mathematical model under the dq coordinate system in the step one, and discretizing to obtain a standard identification form of the doubly-fed motor under the dq coordinate system;
step four, sampling in real time to obtain three-phase stator voltage u when the double-fed motor operatesA、uB、uCThree-phase stator current iA、iB、iCThree-phase rotor voltage ua、ub、ucThree-phase rotor current ia、ib、icAngular frequency ω corresponding to the motormFor stator voltage and currentClark conversion and Park conversion are carried out on the sampling value to respectively obtain the stator voltage U under the dq coordinate systemsd、UsqStator current isd、isqRotor voltage Urd、UrqRotor current ird、irqAngular stator current frequency omega1Calculating to obtain the angular frequency omega of the rotor current for a fixed value2；
Step five, obtaining the stator voltage U in the step foursd、UsqStator current isd、isqRotor voltage Urd、UrqRotor current ird、irqSubstituting the standard identification form obtained in the step three to obtain the parameter to be measured: stator resistance RsStator inductance LsStator-rotor mutual inductance Lm；
And step six, repeating the step four and the step five, iteratively calculating the parameter to be measured, and continuously updating and calculating to obtain a new parameter value to be measured.
Specifically, in the first step, assuming that the stator current is positive by outflow and the rotor current is positive by inflow, the mathematical model of the motor is subjected to Clark transformation and Park transformation, and the mathematical model of the doubly-fed motor under a synchronous rotation dq axis coordinate system can be obtained as follows: the formula I is as follows:
Figure GDA0003133402600000021
the formula II is as follows:
Figure GDA0003133402600000022
wherein, Usd、UsqRespectively, dq-axis stator voltage component, Urd、UrqDq axis rotor voltage components, respectively; i.e. isd、isqThe dq axis stator current components, respectively; i.e. ird、irqThe dq axis stator current components, respectively; psisd、ψsqDq-axis stator flux components, respectively; psird、ψrqFlux linkages of rotor of dq-axis, respectivelyA component; omega1For stator current angular frequency, i.e. grid line frequency 50Hz, omega2Is the rotor current angular frequency; d is a differential operator; rrIs the rotor resistance;
the formula III is as follows:
Figure GDA0003133402600000023
the formula four is as follows:
Figure GDA0003133402600000024
wherein L ismThe stator and the rotor are mutually inducted; l issIs a stator inductance; l isrIs the rotor inductance.
Specifically, the second step comprises the following steps:
step 2.1, obtaining a random gradient algorithm according to a linear regression model;
in the case of a linear regression model,
the formula five is as follows:
Figure GDA0003133402600000025
wherein y (t) is an output vector,
Figure GDA0003133402600000026
is an information vector, theta is a parameter to be identified, v (t) is e R1Is a noise vector;
let the objective function be
Figure GDA0003133402600000027
Wherein the norm of X is defined as | | | X | | non-volatile memory2＝tr[XXT]，tr[X]A trace representing X; minimizing J (theta) according to a gradient search principle to obtain a random gradient algorithm:
formula six:
Figure GDA0003133402600000028
the formula seven:
Figure GDA0003133402600000031
the formula eight:
Figure GDA00031334026000000313
wherein
Figure GDA0003133402600000032
Respectively estimated values of theta at the current time and the last time, e (t) is a single innovation,
Figure GDA0003133402600000033
is a convergence factor;
step 2.2, introducing an innovation length p, and expanding the original single innovation quantity e (t) to a multi-innovation vector with the data length p, namely obtaining a multi-innovation random gradient algorithm according to a random gradient algorithm;
wherein, the original single innovation vector E (t) is expanded to a multiple innovation vector E (p, t) with the data length p, and the following results are obtained:
the formula is nine:
Figure GDA0003133402600000034
formula ten:
Figure GDA0003133402600000035
formula eleven:
Y(p,t)＝[y(t) y(t-1) … y(t-p+1)]∈R1×p
wherein
Figure GDA0003133402600000036
A value representing a past time;
obtaining a multi-innovation random gradient algorithm according to a random gradient algorithm:
equation twelve:
Figure GDA0003133402600000037
formula thirteen:
Figure GDA0003133402600000038
the formula fourteen:
Figure GDA0003133402600000039
step 2.3, adding a variable forgetting factor FF (t) into the multi-innovation random gradient identification method to obtain a variable forgetting factor multi-innovation random gradient algorithm identification expression;
equation fifteen:
Figure GDA00031334026000000310
wherein (FF1, FF2) is the range of variation of FF (t), δmThe maximum error is allowed, and when the error delta (t) > t is defined, delta (t) is taken as delta; in that
Figure GDA00031334026000000311
When the system parameter is identified, δ (t) is 0.2 δ, where δ is a norm of an error between the system parameter and a real parameter, i.e., an identification error of the system parameter;
obtaining a variable forgetting factor multi-innovation random gradient algorithm identification expression:
the formula sixteen:
Figure GDA00031334026000000312
specifically, the third step comprises the following steps:
step 3.1, transforming the mathematical model under the dq coordinate system to obtain a standard identification form of the doubly-fed motor under the dq coordinate system;
under a rotating coordinate system, substituting a stator and rotor magnetic chain equation formula three and a stator and rotor magnetic chain equation formula four into a voltage equation formula one and a voltage equation formula two, wherein a mathematical voltage equation of the doubly-fed motor under a synchronous rotating dq coordinate system is as follows:
the formula seventeen:
Figure GDA0003133402600000041
eighteen formulas:
Figure GDA0003133402600000042
rewriting the seventeenth formula and the eighteen formula into a matrix form as follows:
the formula is nineteen:
Figure GDA0003133402600000043
the formula twenty:
Figure GDA0003133402600000044
step 3.2, taking the sampling period as T, and discretizing differential operators D in the formula nineteen and the formula twenty to obtain a discrete identification standard form under a dq coordinate system of the doubly-fed motor, wherein the identification standard form is as follows:
the formula twenty one:
Figure GDA0003133402600000045
the formula twenty-two:
Figure GDA0003133402600000046
considering only the stator side, the autoregressive model of a doubly-fed machine is:
the formula twenty-three:
Figure GDA0003133402600000047
wherein y (k) ═ Usd(k) Usq(k)]T
Figure GDA0003133402600000048
θ＝[Ls Lm Rs]T。
The invention has the advantages that: according to the method, a mathematical model of the doubly-fed motor under a synchronous rotation dq coordinate system is transformed into a standard form of identification through a series of transformation deduction, so that the identification of the doubly-fed motor parameters by adopting a variable forgetting factor multi-information random gradient method becomes possible; compared with the traditional random gradient algorithm and the least square method, the multi-information random gradient method enables the data to be fully utilized, improves the convergence speed, and avoids the condition of large calculation amount caused by the calculation of covariance in the least square method; the variable forgetting factor is added, so that the convergence accuracy of the identification parameters can be improved, and the forgetting factor is small enough under the non-stationary condition, so that the algorithm can quickly track the local trend of the non-stationary signal; under the steady state condition, the forgetting factor is gradually increased to a proper value so as to reduce the estimation error of the parameter; the motor parameters are identified and updated on line in real time, so that accurate parameters of the motor are obtained in real time, the precision of a control model is improved, and the control performance of the system is improved.
Drawings
Fig. 1 is a DFIG space vector diagram of stator flux linkage orientation.
FIG. 2 is a flow chart of motor parameter identification.
Detailed Description
In order that the objects, aspects and advantages of the present invention will become more apparent, the following detailed description and process of the present invention are given in conjunction with the accompanying drawings.
The invention provides a doubly-fed motor parameter identification method based on a variable forgetting factor multi-information random gradient method aiming at the characteristic of parameter change in the motor operation process, based on a multi-information identification theory and a random gradient identification method and adding a variable forgetting factor. According to the method, nonlinear strong coupling of the doubly-fed motor is considered, a stator flux linkage oriented vector control technology is adopted, a doubly-fed motor vector control system is set up to collect data, a standard form of a motor parameter identification model under a dq coordinate system is deduced, and motor inductance and resistance parameters are calculated iteratively according to the method.
As shown in fig. 2, the whole doubly-fed machine parameter online identification process is as follows: establishing a mathematical model; determining a standard form of a multi-innovation random gradient method, performing real-time sampling, performing data conversion, performing iterative calculation on parameters to be detected, and repeating the steps of performing real-time sampling, performing data conversion, and performing iterative calculation on the parameters to be detected to obtain an identification result. The specific method comprises the following steps:
step 1, establishing a mathematical model of the doubly-fed motor under a dq synchronous coordinate system.
As shown in FIG. 1, in the figure, αs、βsIs the stator-side component, α, in the stationary framer、βrIs the rotor-side component, θ, in the stationary coordinate systemsIs stator flux angle, θrIs the electrical angle of the rotor, thetaslipIs the slip angle, psisIs the stator flux linkage. Assuming that a stator adopts a generator convention, a current takes an outflow as positive, a rotor adopts a motor convention, a current takes an inflow as positive, and a mathematical model of the doubly-fed motor under a synchronous rotation dq axis coordinate system can be obtained by performing Clark conversion and Park conversion on a mathematical model of the motor:
the formula I is as follows:
Figure GDA0003133402600000051
the formula II is as follows:
Figure GDA0003133402600000052
wherein, Usd、UsqRespectively, dq-axis stator voltage component, Urd、UrqDq axis rotor voltage components, respectively; i.e. isd、isqThe dq axis stator current components, respectively; i.e. ird、irqThe dq axis stator current components, respectively; psisd、ψsqDq-axis stator flux components, respectively; psird、ψrqDq axis rotor flux components, respectively; omega1Is the stator current angular frequency; i.e. the power frequency of the grid 50hHz, omega2Is the rotor current angular frequency; d is a differential operator.
The formula III is as follows:
Figure GDA0003133402600000053
the formula four is as follows:
Figure GDA0003133402600000054
wherein L ismThe stator and the rotor are mutually inducted; l issIs a stator inductance; l isrIs the rotor inductance.
Step 2, improving the random gradient method, introducing an innovation length p, expanding the original single innovation quantity e (t) to a multi-innovation vector with the data length p, namely obtaining a multi-innovation random gradient algorithm according to the random gradient algorithm; and adding a variable forgetting factor into the multi-innovation random gradient identification method to obtain a variable forgetting factor multi-innovation random gradient algorithm identification expression.
Step 2.1, obtaining a random gradient algorithm according to a linear regression model;
for the linear regression model, equation five:
Figure GDA0003133402600000055
wherein y (t) is an output vector,
Figure GDA0003133402600000056
is an information vector, theta is a parameter to be identified, v (t) is e R1Is a noise vector.
Let the objective function be
Figure GDA0003133402600000061
Wherein the norm of X is defined as | | | X | | non-volatile memory2＝tr[XXT]，tr[X]Represents the trace of X. Minimizing J (theta) according to a gradient search principle to obtain a random gradient algorithm:
formula six:
Figure GDA0003133402600000062
the formula seven:
Figure GDA0003133402600000063
the formula eight:
Figure GDA0003133402600000064
wherein
Figure GDA0003133402600000065
Are estimated values of theta at the current time and the last time respectively, and e (t) is an innovation.
Step 2.2, introducing an innovation length p, and expanding the original single innovation quantity e (t) to a multi-innovation vector with the data length p, namely obtaining a multi-innovation random gradient algorithm according to a random gradient algorithm;
compared with the least square algorithm, the random gradient algorithm does not need to calculate a covariance matrix, so that the calculated amount is reduced, but the random gradient algorithm has low convergence speed due to low data utilization rate, and in order to improve the convergence speed of parameter estimation of the random gradient algorithm, an innovation length p is introduced, the original single innovation amount e (t) is expanded to a multi-innovation vector with the data length p, so that the data utilization rate of each calculation is improved, and the method is obtained:
the formula is nine:
Figure GDA0003133402600000066
formula ten:
Figure GDA0003133402600000067
formula eleven:
Y(p,t)＝[y(t) y(t-1) … y(t-p+1)]∈R1×p
wherein
Figure GDA0003133402600000068
A value representing a past time.
Obtaining a multi-innovation random gradient algorithm according to a random gradient algorithm:
equation twelve:
Figure GDA0003133402600000069
formula thirteen:
Figure GDA00031334026000000610
the formula fourteen:
Figure GDA00031334026000000611
step 2.3, adding a variable forgetting factor into the multi-innovation random gradient identification method to obtain a variable forgetting factor multi-innovation random gradient algorithm identification expression;
in order to overcome data saturation and reduce the influence of old data, a forgetting factor can be added into the multi-innovation random gradient method. However, a fixed forgetting factor sometimes cannot meet the requirement, and under the non-stationary condition, the FF is usually expected to be small enough, so that the algorithm can quickly track the local trend of the non-stationary signal; in steady state conditions, it is desirable that FF can be gradually increased to an appropriate value to reduce the estimation error of the parameter; therefore, a variable forgetting factor is proposed, and the size of the forgetting factor is updated by detecting the error of the system.
Equation fifteen:
Figure GDA0003133402600000071
wherein (FF1, FF2) is the variation range of FF, δmThe allowable maximum error is defined, and when delta (t) > t, delta (t) is taken as delta; in that
Figure GDA0003133402600000072
When the system parameter is identified, δ (t) is 0.2 δ, where δ is a norm of an error between the system parameter and a real parameter, i.e., an identification error of the system parameter.
Adding a variable forgetting factor into the multiple innovation random gradient identification method to obtain a variable forgetting factor multiple innovation random gradient algorithm formula sixteen:
Figure GDA0003133402600000073
and 3, transforming the mathematical model under the dq coordinate system in the step one, taking the sampling period as T, and performing discretization processing to obtain the discrete identification standard form under the dq coordinate system of the doubly-fed motor.
And 3.1, transforming the mathematical model under the dq coordinate system to obtain a standard identification form of the doubly-fed motor under the dq coordinate system.
In system identification, linearity mainly refers to whether the output of a system is linear to an unknown parameter in a system model, so that the method needs to deform a mathematical model of the doubly-fed motor under a synchronous rotation dq coordinate system into a standard form of parameter identification. Under a rotating coordinate system, substituting a stator and rotor magnetic chain equation formula three and a stator and rotor magnetic chain equation formula four into a voltage equation formula one and a voltage equation formula two, wherein a mathematical voltage equation of the doubly-fed motor under a synchronous rotating dq coordinate system is as follows:
the formula seventeen:
Figure GDA0003133402600000074
eighteen formulas:
Figure GDA0003133402600000075
rewriting the seventeenth formula and the eighteen formula into a matrix form as follows:
the formula is nineteen:
Figure GDA0003133402600000076
the formula twenty:
Figure GDA0003133402600000077
step 3.2, taking the sampling period as T, and discretizing differential operators D in the formula nineteen and the formula twenty to obtain a discrete identification standard form under a dq coordinate system of the doubly-fed motor, wherein the identification standard form is as follows:
the formula twenty one:
Figure GDA0003133402600000078
the formula twenty-two:
Figure GDA0003133402600000079
the autoregressive model of a doubly-fed machine is (considering only the stator side):
the formula twenty-three:
Figure GDA0003133402600000081
wherein y (k) ═ Usd(k) Usq(k)]T
Figure GDA0003133402600000082
θ＝[Ls Lm Rs]T
Step four, sampling in real time to obtain three-phase stator voltage u when the double-fed motor operatesA、uB、uCThree-phase stator current iA、iB、iCThree-phase rotor voltage ua、ub、ucThree-phase rotor current ia、ib、icAngular frequency ω corresponding to the motormClark conversion and Park conversion are carried out on the voltage and current sampling values of the stator and the rotor to respectively obtain stator voltage U under the dq coordinate systemsd、UsqStator current isd、isqRotor voltage Urd、UrqRotor current ird、irqAngular stator current frequency omega1Calculating to obtain the angular frequency omega of the rotor current for a fixed value2；
Step five, obtaining the stator voltage U in the step foursd、UsqStator current isd、isqRotor voltage Urd、UrqRotor current ird、irqSubstituting the standard identification form obtained in the step three to obtain the parameter to be measured: stator resistance RsStator inductance LsStator-rotor mutual inductance Lm；
And step six, repeating the step four and the step five, iteratively calculating the parameter to be measured, and continuously updating and calculating to obtain a new parameter value to be measured. The iteration number can be set manually according to actual conditions, and in the embodiment, the iteration number is set to be 5000.
In conclusion, the invention provides the doubly-fed motor parameter identification method based on the variable forgetting factor multi-information random gradient method aiming at the parameter change characteristics in the motor operation process and based on the multi-information identification theory and the random gradient identification method and adding the variable forgetting factor. According to the method, nonlinear strong coupling of the doubly-fed motor is considered, a stator flux linkage oriented vector control technology is adopted, a doubly-fed motor vector control system is set up to collect data, a standard form of a motor parameter identification model under a dq coordinate system is deduced, and motor inductance and resistance parameters are calculated iteratively according to the method. The method is suitable for the double-fed motor in two states of electric and power generation, can identify the motor stator parameters on line, improves the control model precision and improves the control performance.

Claims (2)

1. A doubly-fed motor parameter online identification method is characterized by comprising the following steps:
step one, establishing a mathematical model of the doubly-fed motor under a synchronous rotation dq coordinate system;
step two, improving the random gradient method, introducing an innovation length p, expanding the original single innovation quantity e (t) to a multi-innovation vector with the data length p, namely obtaining a multi-innovation random gradient algorithm according to the random gradient algorithm; adding a variable forgetting factor into the multi-innovation random gradient identification method to obtain a variable forgetting factor multi-innovation random gradient algorithm identification expression;
step three, transforming the mathematical model under the dq coordinate system in the step one, and discretizing to obtain a standard identification form of the doubly-fed motor under the dq coordinate system;
step four, sampling in real time to obtain three-phase stator voltage u when the double-fed motor operatesA、uB、uCThree-phase stator current iA、iB、iCThree-phase rotor voltage ua、ub、ucThree-phase rotor current ia、ib、icAngular frequency ω corresponding to the motormClark conversion and Park conversion are carried out on the voltage and current sampling values of the stator and the rotor to respectively obtain stator voltage U under the dq coordinate systemsd、UsqStator current isd、isqRotor voltage Urd、UrqRotor current ird、irqAngular stator current frequency omega1Calculating to obtain the angular frequency omega of the rotor current for a fixed value2；
Step five, obtaining the stator voltage U in the step foursd、UsqStator current isd、isqRotor voltage Urd、UrqRotor current ird、irqSubstituting the standard identification form obtained in the step three to obtain the parameter to be measured: stator resistance RsStator inductance LsStator-rotor mutual inductance Lm；
Step six, repeating the step four and the step five, iteratively calculating the parameter to be measured, and continuously updating and calculating to obtain a new parameter value to be measured;
assuming that the stator current is positive when flowing out and the rotor current is positive when flowing in, and performing Clark transformation and Park transformation on a motor mathematical model to obtain the mathematical model of the doubly-fed motor under a synchronous rotation dq axis coordinate system as follows:
the formula I is as follows:
Figure FDA0003133402590000011
the formula II is as follows:
Figure FDA0003133402590000012
wherein psisd、ψsqDq-axis stator flux components, respectively; psird、ψrqDq axis rotor flux components, respectively; omega1The stator current angular frequency is the power frequency of the power grid of 50 Hz; d is a differential operator; rrIs the rotor resistance;
the formula III is as follows:
Figure FDA0003133402590000013
the formula four is as follows:
Figure FDA0003133402590000014
wherein L ismThe stator and the rotor are mutually inducted; l issIs a stator inductance; l isrIs a rotor inductance;
the second step comprises the following steps:
step 2.1, obtaining a random gradient algorithm according to a linear regression model;
in the case of a linear regression model,
the formula five is as follows:
Figure FDA0003133402590000015
wherein y (t) is an output vector,
Figure FDA0003133402590000016
is an information vector, theta is a parameter to be identified, v (t) is e R1Is a noise vector;
let the objective function be
Figure FDA0003133402590000017
Wherein the norm of X is defined as | | | X | | non-volatile memory2＝tr[XXT]，tr[X]A trace representing X; minimizing J (theta) according to a gradient search principle to obtain a random gradient algorithm:
formula six:
Figure FDA0003133402590000021
the formula seven:
Figure FDA0003133402590000022
the formula eight:
Figure FDA0003133402590000023
wherein
Figure FDA0003133402590000024
Respectively estimated values of theta at the current time and the last time, e (t) is a single innovation,
Figure FDA0003133402590000025
is a convergence factor;
step 2.2, introducing an innovation length p, and expanding the original single innovation quantity e (t) to a multi-innovation vector with the data length p, namely obtaining a multi-innovation random gradient algorithm according to a random gradient algorithm;
wherein, the original single innovation vector E (t) is expanded to a multiple innovation vector E (p, t) with the data length p, and the following results are obtained:
the formula is nine:
Figure FDA0003133402590000026
formula ten:
Figure FDA0003133402590000027
formula eleven:
Y(p,t)＝[y(t) y(t-1) … y(t-p+1)]∈R1×p
wherein y (t-i) is,
Figure FDA0003133402590000028
p-1 represents a value of past time;
obtaining a multi-innovation random gradient algorithm according to a random gradient algorithm:
equation twelve:
Figure FDA0003133402590000029
formula thirteen:
Figure FDA00031334025900000210
the formula fourteen:
Figure FDA00031334025900000211
step 2.3, adding a variable forgetting factor FF (t) into the multi-innovation random gradient identification method to obtain a variable forgetting factor multi-innovation random gradient algorithm identification expression;
equation fifteen:
Figure FDA00031334025900000212
wherein (FF1, FF2) is the range of variation of FF (t), δmThe maximum error is allowed, and when the error delta (t) > t is defined, delta (t) is taken as delta; in that
Figure FDA00031334025900000213
When the system parameter is identified, δ (t) is 0.2 δ, where δ is a norm of an error between the system parameter and a real parameter, i.e., an identification error of the system parameter;
obtaining a variable forgetting factor multi-innovation random gradient algorithm identification expression:
the formula sixteen:
Figure FDA0003133402590000031
2. the method for online identifying the parameters of the doubly-fed motor of claim 1, wherein the third step comprises the following steps:
step 3.1, transforming the mathematical model under the dq coordinate system to obtain a standard identification form of the doubly-fed motor under the dq coordinate system;
under a rotating coordinate system, substituting a stator and rotor magnetic chain equation formula three and a stator and rotor magnetic chain equation formula four into a voltage equation formula one and a voltage equation formula two, wherein a mathematical voltage equation of the doubly-fed motor under a synchronous rotating dq coordinate system is as follows:
the formula seventeen:
Figure FDA0003133402590000032
eighteen formulas:
Figure FDA0003133402590000033
rewriting the seventeenth formula and the eighteen formula into a matrix form as follows:
the formula is nineteen:
Figure FDA0003133402590000034
the formula twenty:
Figure FDA0003133402590000035
step 3.2, taking the sampling period as T, and discretizing differential operators D in the formula nineteen and the formula twenty to obtain a discrete identification standard form under a dq coordinate system of the doubly-fed motor, wherein the identification standard form is as follows:
the formula twenty one:
Figure FDA0003133402590000036
the formula twenty-two:
Figure FDA0003133402590000037
considering only the stator side, the autoregressive model of a doubly-fed machine is:
the formula twenty-three:
Figure FDA0003133402590000038
wherein y (k) ═ Usd(k) Usq(k)]T
Figure FDA0003133402590000039
θ＝[Ls Lm Rs]T。
CN201811156781.8A 2018-09-30 2018-09-30 Online identification method for parameters of doubly-fed motor Active CN109194225B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811156781.8A CN109194225B (en) 2018-09-30 2018-09-30 Online identification method for parameters of doubly-fed motor

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811156781.8A CN109194225B (en) 2018-09-30 2018-09-30 Online identification method for parameters of doubly-fed motor

Publications (2)

Publication Number Publication Date
CN109194225A CN109194225A (en) 2019-01-11
CN109194225B true CN109194225B (en) 2021-08-20

Family

ID=64946669

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811156781.8A Active CN109194225B (en) 2018-09-30 2018-09-30 Online identification method for parameters of doubly-fed motor

Country Status (1)

Country Link
CN (1) CN109194225B (en)

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109524987B (en) * 2018-11-19 2022-02-18 贵州大学 Constant voltage control-based doubly-fed wind generator mutual inductance parameter identification method
CN110286332B (en) * 2019-06-17 2021-06-22 南通大学 Electric vehicle power battery SOC estimation method based on multiple innovation theory
CN110333667A (en) * 2019-08-14 2019-10-15 大连海事大学 A kind of Simulation of ship motion and control method based on non-linear new breath identification model
CN112350638A (en) * 2020-10-12 2021-02-09 北京自动化控制设备研究所 Servo motor parameter identification method based on nonlinear optimization
CN113659906B (en) * 2021-08-31 2023-07-07 北京信息科技大学 Online identification method for unknown motor parameters
CN115343570B (en) * 2022-07-21 2023-05-23 国网湖北省电力有限公司电力科学研究院 Online identification method and device for power grid impedance

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102611380A (en) * 2012-03-09 2012-07-25 哈尔滨工业大学 Online identification method for parameters of double-fed motor
CN103795071A (en) * 2014-01-24 2014-05-14 武汉大学 Wide-area damping control method based on identification of power system model
CN103825521A (en) * 2014-02-25 2014-05-28 河海大学 Method for identifying parameters of double-fed wind turbine generator driving system and generator
US9391554B2 (en) * 2010-08-25 2016-07-12 University Of Alabama Control of a permanent magnet synchronous generator wind turbine
CN106487297A (en) * 2016-11-24 2017-03-08 北京邮电大学 A kind of PMSM parameter identification method based on covariance matching Unscented kalman filtering algorithm

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9391554B2 (en) * 2010-08-25 2016-07-12 University Of Alabama Control of a permanent magnet synchronous generator wind turbine
CN102611380A (en) * 2012-03-09 2012-07-25 哈尔滨工业大学 Online identification method for parameters of double-fed motor
CN103795071A (en) * 2014-01-24 2014-05-14 武汉大学 Wide-area damping control method based on identification of power system model
CN103825521A (en) * 2014-02-25 2014-05-28 河海大学 Method for identifying parameters of double-fed wind turbine generator driving system and generator
CN106487297A (en) * 2016-11-24 2017-03-08 北京邮电大学 A kind of PMSM parameter identification method based on covariance matching Unscented kalman filtering algorithm

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Parameter identification for PMSM based on varying forgetting factor multi-innovation stochastic gradient identification algorithm;Wu Dinghui et al;《2017 36th Chinese Control Conference (CCC)》;20170728;第2027-2032页 *
基于遗忘因子多新息随机梯度算法PMSM参数辨识;张建宇 等;《微特电机》;20161130;第44卷(第11期);第66-69页 *

Also Published As

Publication number Publication date
CN109194225A (en) 2019-01-11

Similar Documents

Publication Publication Date Title
CN109194225B (en) Online identification method for parameters of doubly-fed motor
CN102611380B (en) Online identification method for parameters of double-fed motor
CN107451364B (en) Method for identifying equivalent parameters of DFIG wind power plant
CN108092567A (en) A kind of Speed control of permanent magnet synchronous motor system and method
CN102176653A (en) Method for observing rotary speed of induction motor of Kalman filter with index fading factor
CN112818491A (en) Wind power plant aggregation equivalent modeling method based on principal component analysis and clustering algorithm
CN113285647A (en) Permanent magnet synchronous motor feedback adjustment method and device and permanent magnet synchronous motor
CN108365785A (en) A kind of asynchronous machine repetition forecast Control Algorithm
Moon et al. Nonlinear neural-network modeling of an induction machine
CN115249980B (en) New energy station impedance identification method and system driven by combination of data and knowledge
Yan et al. Torque estimation and control of PMSM based on deep learning
CN109946978A (en) A kind of servo-system fractional model discrimination method considering time delay process
CN116306306A (en) Micro-grid inertia constant estimation method based on African bald-Condition algorithm
CN116738636A (en) Multi-machine equivalent method of doubly-fed wind power station considering impedance characteristics and synchronization mechanism
Bohari et al. Speed tracking of indirect field oriented control induction motor using neural network
CN115392133A (en) Wind power plant optimal clustering equivalence method adopting Gaussian mixture model
Pawlus et al. Parameters identification of induction motor dynamic model for offshore applications
CN111179110B (en) Virtual power plant variable order aggregation equivalent robust dynamic model modeling method and device
CN114465280A (en) Dynamic equivalent modeling method for new energy grid-connected system
CN112630596A (en) Comprehensive diagnosis method for open-circuit fault of IGBT device of wind power converter
CN113794198A (en) Method, device, terminal and storage medium for suppressing broadband oscillation
Kamwa et al. Sensorless ANN-based speed estimation of synchronous generators: Improved performance through physically motivated pre-filters
CN111181461A (en) DFIG stator flux observer and method based on resistance online identification
CN111181455A (en) Self-adaptive tracking control method and system for maximum power point of doubly-fed wind generator
CN115940248B (en) Wind-storage combined power generation system damping evaluation method based on EMD and dissipated energy

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant