CN114172201B - Modular direct-driven fan control model structure identification method - Google Patents

Abstract

The invention provides a modular direct-driven fan control model structure identification method, which comprises the following steps: dividing the control function of the direct-drive fan into a plurality of control modules, wherein the control modules comprise an active power control module, a reactive power output limiting module, a reactive power control module and a current limiting module; step two, constructing a direct-drive fan control model, adjusting input current, and identifying a current limiting link of a current limiting module; reactive power is regulated, and a reactive power limiting link of a reactive power output limiting module is identified; step three, the orders of the active power control module and the reactive power control module are identified by using an MSVC criterion, and the parameters of the active power control module and the reactive power control module are identified by using an N4SID criterion respectively; fourth, the active power control model obtained through identification

And reactive power control model

Sequentially verifying; the method can accurately and reliably identify the orders of the control modules of the direct-drive fan.

Description

Modular direct-driven fan control model structure identification method

Technical Field

The invention relates to the technical field of new energy power generation parameter identification, in particular to a modular direct-driven fan control model structure identification method.

Background

The direct-drive permanent magnet wind driven generator is one of the main stream models in the field of wind power generation, the wind driven generator is directly driven by the wind driven generator, a speed increasing gear box is omitted, and the permanent magnet motor does not need electric excitation, so that the control is simpler, and the direct-drive permanent magnet wind driven generator has the advantages of good control effect, high reliability and the like, and gradually becomes a research focus of people. Meanwhile, the large-scale wind power is connected into the power grid to bring a plurality of challenges to the safe and stable operation of the power system, the influence on peak shaving, voltage and frequency of the power grid is increasingly remarkable, and the situation of the power grid is taken up seriously. In recent years, serious problems such as fan trawl and equipment damage are caused by the occurrence of major subsynchronous oscillation time of a wind power plant. The oscillation frequency ranges from a few hertz to a few kilohertz, and the reason for the oscillation is closely related to the electromechanical system of the fan, the electromagnetic system, the power grid parameters and other factors.

The electromechanical transient model capable of reflecting the actual characteristics of new energy power generation is established, so that basic conditions can be provided for grasping interaction influence mechanisms of new energy power generation and a power grid, the operation mode and control measures of the power grid are supported for optimization, and the full consumption of new energy is ensured. In order to ensure the consistency of simulation results and actual characteristics and also to give consideration to higher simulation efficiency, higher requirements on a new energy simulation model are required.

In the direct-driven fan parameter modeling and simulation, because the power generation control parameters are numerous, and simulation research needs to be conducted on model selection and parameter assumption, which implies diversity and uncertainty of models and parameters, parameter identification needs to be conducted on the models in order to obtain model parameters with more engineering universality on the basis of actual test data with enough units so as to ensure the universality of the models. The structure of the direct-driven fan needs to be identified before the parameter identification is carried out on the direct-driven fan, the direct-driven fan in reality can be regarded as an infinite-order system, and when the simulation is carried out, the optimal order of the system to be identified is determined, so that the direct-driven fan can be better fitted with the real system without paying excessive calculation cost.

Disclosure of Invention

The invention provides a modular direct-drive fan control model structure identification method which can accurately and reliably identify the orders of all control modules of a direct-drive fan.

A modular direct-driven fan control model structure identification method comprises the following steps:

dividing the control function of the direct-drive fan into a plurality of control modules, wherein the control modules comprise an active power control module, a reactive power output limiting module, a reactive power control module and a current limiting module;

step two, constructing a direct-drive fan control model, adjusting input current, and identifying a current limiting link of a current limiting module; reactive power is regulated, and a reactive power limiting link of a reactive power output limiting module is identified;

step three, the orders of the active power control module and the reactive power control module are identified by using an MSVC criterion, and the parameters of the active power control module and the reactive power control module are identified by using an N4SID criterion respectively;

fourth, the active power control model obtained through identification

And reactive power control model->

And sequentially verifying.

The input of the active power control module in the first step is the active power reference value P of the fan _Wref Angular velocity omega of fan rotor _gen Dc bus voltage u _dc Output net side current value d-axis component i of current amplitude limiting link _gd The output of the module is the active power P output by the fan; the input of the reactive power control module is the DC bus voltage u _dc Output net side current value q axis component i with current amplitude limiting link _gq Reactive power reference value Q of fan _ref The output of the module is the reactive power Q output by the fan; the input of the current limiting module is a network side current value i _g And the active power P is output by the active power control module; the input of the reactive power limiting module is the active power P output by the active power control module and the reactive power Q output by the reactive power control module, and the output is the Q after limiting ₁ 。

The third step comprises the following steps;

step A1, collecting an active power control model G ₁ Input signal u of (2) ₁ And output signal y ₁ And reactive power control model G ₂ Input signal u of (2) ₂ And output signal y ₂ ；

Step A2, using the input signal u ₁ Active power control model obtained by excitation identification

Obtain the output signal +.>

By input signal u ₂ Reactive power control model obtained by excitation identification>

Obtain the output signal +.>

Step A3, respectively y ₁ And (3) with

And y ₂ And->

Comparing, and if the fitting degree between the two is high, identifying to obtain a reliable model; if the fitting degree of the two is larger, the model obtained by identification is unreliable; judging by using a fitting ratio index FR, wherein the fitting ratio index is defined as:

where N is the number of samples acquired, the closer FR to 1 indicates the better fit, and otherwise indicates the worse fit.

The second step comprises the following steps of;

step S1: constructing a direct-drive fan control model;

step S2: collecting measurable input and output signals of a control model and carrying out per unit processing;

step S3: and calculating the undetectable input and output parameters according to the input and output parameters of the control module.

In the third step, MSVC criterion is adopted to identify the system order of the direct drive fan and N4SID algorithm is utilized to identify system parameters, firstly, a state space model of a discrete form of the system is established, and the state space model is expressed as a formula

x _k+1 ＝Ax _k +Bu _k +ω _k

y _k ＝Cx _k +Du _k +v _k A second formula;

in which x is _k ∈R ⁿ ，y _k ∈R ^m ，u _k ∈R ^r The input vector, the output vector and the state vector at the moment k of the system are respectively; a epsilon R ^n×n 、B∈R ^n×m 、C∈R ^l×n 、D∈R ^l×m Is a system matrix omega _k And v _k Representing measurement noise and process noise, respectively;

the identifying includes the following steps;

step S41, collecting input and output signals with the sequence of { y }, respectively ₀ ，y ₁ ，...，y _M }、{u ₀ ，u ₁ ，...，u _M }，

Wherein y is _k ∈R ^l×1 ，u _k ∈R ^m×1 ，k＝0，1，...，M；

Step S42, constructing Hankel matrixes of past input, future input, past output and future output, and expressing the matrix as a formula:

in the same way, use u _k Construction of matrix U _p ∈R ^mi×j ，

U _f ∈R ^mi×j

Step S43, defining a matrix W _p And W is _p+ ：

Matrix Y _f Along matrix U _f In matrix W _p Matrix O obtained by oblique projection _i ：

Formula nine;

matrix Y _f In matrix U _f And W is _p Matrix Z obtained by orthographic projection on new matrix _i ：

Matrix array

In matrix->

And->

Matrix Z obtained by orthographic projection on new matrix _i+1 ：

In the above, symbol () ⁺ Representing a matrix Moore-Penrose pseudo-inverse

Step S44: for matrix O _i Singular value decomposition, SVD, is performed, SVD is Singular Value Decomposition:

wherein U is ₁ ∈R ^li×n ，U ₂ ∈R ^li×(li-n) ，S ₁ ∈R ^n×n ，S ₂ ∈R ^{(Ii-n)×(j-n)} ，V ₁ ^T ∈R ^n×j ，V ₂ ^T ∈R ^(j-n)×j ；

Step S45: calculating and expanding observable matrix gamma _i ：

Γ _i ＝U ₁ S ₁ ^1/2 Formula thirteen;

Γ _i-1 ＝fist(i-1)l rows ofΓ _i formula fourteen;

step S46: solving the linear system of equations yields matrices A, C and K:

wherein Y is _i|i ＝[y _i y _i+1 … y _i+j-1 ]∈R ^l×j ，

Is with Z _i And U _f Orthogonal noise terms;

step S47: the model order is identified by using MSVC criteria, namely Modified Singular Value Criterion:

where T is the number of matrix elements of the future output block Hankel,

the square of matrix 2 norms formed by singular value identification values is represented by C (T), which is a function of T;

step S48: using least squares to pass through matrix A, C, Γ _i And Γ _i-1 The matrices B and D are solved.

The invention provides a modular direct-driven fan control model structure identification method. Firstly, exploring a control model structure of a modularized direct-driven fan, establishing a wind power generation related environmental factor random model, and adjusting input records and output of a limiting link to identify the limiting link; and then, a subspace model identification method N4SID method is utilized to identify an active power control model and a reactive power control model of the fan, the method comprises the steps of collecting input and output data of the active power control model and the reactive power control model, preprocessing the data, respectively carrying out order identification on the active power model and the reactive power model of the direct-drive fan by adopting an MSVC criterion, and comparing the control output obtained by identification with the output of the original control model to ensure the suitability of the order of the identification model. The method can identify the structure of the modularized direct-driven fan control model to obtain parameters of a current limiting link and a reactive power limiting link, and the order and the parameters of an active power control model and a reactive power control model.

Compared with the prior art, the invention has the following beneficial effects:

the technical scheme provides a modularized direct-driven fan control model structure identification method, which introduces a subspace model identification method into direct-driven fan control model structure identification, directly drives a fan control model, and establishes a wind power generation related environmental factor random model. And processing the data by using the input and output parameters of each control module and adopting an N4SID algorithm, thereby obtaining the order of the control model structure.

The invention combines the actual working conditions, identifies the control model structure based on the general structure of the direct-drive fan, and verifies the identification model. The method has the advantages of high reliability and good universality.

Drawings

The invention is described in further detail below with reference to the attached drawings and detailed description:

FIG. 1 is a schematic flow chart of a judging method in an application embodiment;

FIG. 2 is a schematic diagram of the fitting error of the recognition output result of the active control module in the application embodiment;

fig. 3 is a schematic diagram of the fitting error of the recognition output result of the reactive power control module in the application embodiment.

Detailed Description

As shown in the figure, the method for identifying the control model structure of the modularized direct-driven fan comprises the following steps:

dividing the control function of the direct-drive fan into a plurality of control modules, wherein the control modules comprise an active power control module, a reactive power output limiting module, a reactive power control module and a current limiting module;

step two, constructing a direct-drive fan control model, adjusting input current, and identifying a current limiting link of a current limiting module; reactive power is regulated, and a reactive power limiting link of a reactive power output limiting module is identified;

step three, the orders of the active power control module and the reactive power control module are identified by using an MSVC criterion, and the parameters of the active power control module and the reactive power control module are identified by using an N4SID criterion respectively;

fourth, the active power control model obtained through identification

And reactive power control model->

And sequentially verifying.

The input of the active power control module in the first step is the active power reference value P of the fan _Wref Angular velocity omega of fan rotor _gen Dc bus voltage u _dc Output net side current value d-axis component i of current amplitude limiting link _gd The output of the module is the active power P output by the fan; the input of the reactive power control module is the DC bus voltage u _dc Output net side current value q axis component i with current amplitude limiting link _gq Reactive power reference value Q of fan _ref The output of the module is the reactive power Q output by the fan; the input of the current limiting module is a network side current value i _g And the active power P is output by the active power control module; the input of the reactive power limiting module is the active power P output by the active power control module and the reactive power Q output by the reactive power control module, and the output is the Q after limiting ₁ 。

The third step comprises the following steps;

step A1, collecting an active power control model G ₁ Input signal u of (2) ₁ And output signal y ₁ And reactive power control model G ₂ Input signal u of (2) ₂ And output signal y ₂ ；

Step A2, using the input signal u ₁ Active power control model obtained by excitation identification

Obtain the output signal +.>

By input signal u ₂ Reactive power control model obtained by excitation identification>

Obtain the output signal +.>

Step A3, respectively y ₁ And (3) with

And y ₂ And->

Comparing, and if the fitting degree between the two is high, identifying to obtain a reliable model; if the fitting degree of the two is larger, the model obtained by identification is unreliable; judging by using a fitting ratio index FR, wherein the fitting ratio index is defined as:

where N is the number of samples acquired, the closer FR to 1 indicates the better fit, and otherwise indicates the worse fit.

The second step comprises the following steps of;

step S1: constructing a direct-drive fan control model;

step S2: collecting measurable input and output signals of a control model and carrying out per unit processing;

step S3: and calculating the undetectable input and output parameters according to the input and output parameters of the control module.

In the third step, MSVC criterion is adopted to identify the system order of the direct drive fan and N4SID algorithm is utilized to identify system parameters, firstly, a state space model of a discrete form of the system is established, and the state space model is expressed as a formula

x _k+1 ＝Ax _k +Bu _k +ω _k

y _k ＝Cx _k +Du _k +v _k A second formula;

in which x is _k ∈R ⁿ ，y _k ∈R ^m ，u _k ∈R ^r The input vector, the output vector and the state vector at the moment k of the system are respectively; a epsilon R ^n×n 、B∈R ^n×m 、C∈R ^l×n 、D∈R ^l×m Is a system matrix omega _k And v _k Representing measurement noise and process noise, respectively;

the identifying includes the following steps;

step S41, collecting input and output signals with the sequence of { y }, respectively ₀ ，y ₁ ，...，y _M }、{u ₀ ，u ₁ ，...，u _M -wherein y _k ∈R ^l×1 ，u _k ∈R ^m×1 ，k＝0，1，...，M；

Step S42, constructing Hankel matrixes of past input, future input, past output and future output, and expressing the matrix as a formula:

in the same way, use u _k Construction of matrix U _p ∈R ^mi×j ，

U _f ∈R ^mi×j ，

Step S43, defineMatrix W _p And

matrix Y _f Along matrix U _f In matrix W _p Matrix O obtained by oblique projection _i ：

Matrix Y _f In matrix U _f And W is _p Matrix Z obtained by orthographic projection on new matrix _i ：

Matrix array

In matrix->

And->

Matrix Z obtained by orthographic projection on new matrix _i+1 ：

In the above, symbol () ⁺ Representing a matrix Moore-Penrose pseudoReverse direction

Step S44: for matrix O _i Singular value decomposition, SVD, is performed, SVD is Singular Value Decomposition:

wherein U is ₁ ∈R ^li×n ，U ₂ ∈R ^li×(li-n) ，S ₁ ∈R ^n×n ，S ₂ ∈R ^{(li-n)×(j-n)} ，V ₁ ^T ∈R ^n×j ，V ₂ ^T ∈R ^(j-n)×j ；

Step S45: calculating and expanding observable matrix gamma _i ：

Γ _i ＝U ₁ S ₁ ^1/2 Formula thirteen;

Γ _i-1 ＝furst(i-1)l rows ofΓ _i formula fourteen;

step S46: solving the linear system of equations yields matrices A, C and K:

wherein Y is _i|i ＝[y _i y _i+1 … y _i+j-1 ]∈R ^i×j ，

Is with Z _i And U _f Orthogonal noise terms;

step S47: the model order is identified by using MSVC criteria, namely Modified Singular Value Criterion:

where T is the number of matrix elements of the future output block Hankel,

the square of matrix 2 norms formed by singular value identification values is represented by C (T), which is a function of T;

step S48: using least squares to pass through matrix A, C, Γ _i And Γ _i-1 The matrices B and D are solved.

Examples:

the permanent magnet synchronous generator is used as a fan, is connected into an infinite power grid through an AC-DC-AC full-power converter, has a machine end voltage of 380V, is 0.064 omega of a grid-connected line, has an inductance of 3.2149e-05H, has a system voltage grade of 380V and has a frequency of 50HZ; the maximum power that the fan can emit is 50kW.

The simulation model mainly comprises a permanent magnet synchronous generator, an inverter, a machine side frequency converter control, a network side frequency converter control, a wind speed module, a maximum power tracking module, a wind turbine module and the like.

The control mode of the fan adopts a maximum power tracking mode and adopts a constant power factor 1 grid-connected operation. The simulation time is 1.5s, the wind speed is 5m/s at 0-0.5 s, 7m/s at 0.5-1 s, and 9m/s at 1-1.5 s.

The direct-drive fan control part is divided into four modules of active control, reactive control, current limiting and reactive limiting according to the method. Wherein the active power reference value P _ref Net side current d-axis component i _gd Net side current q-axis component i _gq Reactive power reference value Q _ref Cannot be obtained by direct measurement, but can be obtained by calculation of relevant parameters. The fan control mode is a maximum power tracking mode, and Q is known _ref Is 0; and (3) fixing the pitch angle, and changing the wind speed to draw an optimal power curve of the wind turbine. The optimal power curve corresponds to the formula as follows:

when the wind speed changesDuring conversion, the reference value P of the output power of the generator can be calculated according to the optimal power curve _ref . The fan operates under the fault-free condition, the current and the output reactive power of the network side generally do not exceed the set amplitude, and the amplitude limiting link does not act. The maximum 2 times of input current allowed to flow in the current limiting link is obtained by adjusting the current of the input limiting link to estimate; and the reactive power of the maximum 1 time of the reactive power limiting link is allowed to pass through by adjusting the reactive power of the input limiting link to estimate. Net side current i _g After park transformation, q-axis component i can be obtained _gq And d-axis component i _gd 。

Collecting the input and output signal sequences which can be measured by the active power control module, obtaining the input signal sequences which cannot be measured by calculation, and obtaining the output signal sequences by per unit processing

And input signal sequence { u } _p0 ，u _p1 ，…，u _pM }, wherein

y _pk ∈R ^1×1 ，u _pk ∈R ^4×1 ，k＝0，1，...，M，M＝5000。

The MSVC criterion is adopted to identify the system order of 3, the N4SID algorithm is utilized to identify the system parameters, and the system matrix A, B, C, D is obtained as follows:

C＝[-0.04976 0.0002248 3.249e-07]

D＝[0 0 0 0]

output signal sequence calculated by using the solved parameters

With input signal sequence { u } _p0 ，u _p1 ，...，u _pM And compared, the fitting degree is 94.28 percent, and a better fitting result is obtained.

Collecting an input and output signal sequence which can be measured by a reactive power control module, obtaining an input signal sequence which cannot be measured through calculation, and obtaining an output signal sequence { y) through per unit processing _q0 ，y _q1 ，...，yq _M Sum of input signal sequence u _q0 ，u _q1 ，...，u _qM -wherein y _qk ∈R ^1×1 ，u _qk ∈R ^4×1 ，k＝0，1，...，M，M＝5000。

The MSVC criterion is adopted to identify the system order of 2, the N4SID algorithm is utilized to identify the system parameters, and the system matrix A, B, C, D, K is obtained as follows:

C＝[-0.3817 0.003336]

D＝[0 0 0]

output signal sequence { y } calculated by using the solved parameters _q0 ，y _q1 ，...，y _qM Sequence { u } and input signal _q0 ，u _q1 ，...，u _qM And compared, the fitting degree is 94.19 percent, and a better fitting result is obtained.

Claims (3)

1. A modular direct-driven fan control model structure identification method is characterized by comprising the following steps of: the method comprises the following steps:

dividing the control function of the direct-drive fan into a plurality of control modules, wherein the control modules comprise an active power control module, a reactive power output limiting module, a reactive power control module and a current limiting module;

step two, constructing a direct-drive fan control model, adjusting input current, and identifying a current limiting link of a current limiting module; reactive power is regulated, and a reactive power limiting link of a reactive power output limiting module is identified;

step three, the orders of the active power control module and the reactive power control module are identified by using an MSVC criterion, and the parameters of the active power control module and the reactive power control module are identified by using an N4SID criterion respectively;

fourth, the active power control model obtained through identification

And reactive power control model->

Sequentially verifying;

the second step comprises the following steps of;

step S1: constructing a direct-drive fan control model;

step S2: collecting measurable input and output signals of a control model and carrying out per unit processing;

step S3: calculating the undetectable input and output parameters according to the input and output parameters of the control module;

in the third step, MSVC criterion is adopted to identify the system order of the direct drive fan and N4SID algorithm is utilized to identify system parameters, firstly, a state space model of a discrete form of the system is established, and the state space model is expressed as a formula

x _k+1 ＝Ax _k +Bu _k +ω _k

y _k ＝Cx _k +Du _k +v _k A second formula;

in which x is _k ∈R ⁿ y _k ∈R ^m u _k ∈R ^r The input vector, the output vector and the state vector at the moment k of the system are respectively; a epsilon R ^n×n 、B∈R ^n×m 、C∈R ^l×n 、D∈R ^l×m Is a system matrix omega _k And v _k Representing measurement noise and process noise, respectively;

the identifying includes the following steps;

step S41, collecting input and output signals with the sequence of { y }, respectively ₀ ，y ₁ ，...，y _M }、{u ₀ ，u ₁ ，...，u _M }，

Wherein u is _k ∈R ^l×1 ，u _k ∈R ^m×1 ，k＝0，1，...，M；

Step S42, constructing Hankel matrixes of past input, future input, past output and future output, and expressing the matrix as a formula:

in the same way, use u _k Construction of matrix U _p ∈R ^mi×j ，

U _f ∈R _mi×j ，

Step S43, defining a matrix W _p And

matrix Y _f Along matrix U _f In matrix W _p Matrix O obtained by oblique projection _i ：

Formula nine;

matrix Y _f In matrix U _f And W is _p Matrix Z obtained by orthographic projection on new matrix _i ：

Matrix array

In matrix->

And->

Matrix Z obtained by orthographic projection on new matrix _i+1 ：

In the above, symbol () ⁺ Representing the matrix Moore-Penrose pseudo-inverse,

step S44: moment of alignmentArray O _i Singular Value Decomposition (SVD) is performed:

wherein U is ₁ ∈R ^li×n ，U ₂ ∈R ^li×(li-n) ，S ₁ ∈R ^n×n ，S ₂ ∈R ^{(li-n)×(j-n)} ，V ₁ ^T ∈R ^n×j ，V ₂ ^T ∈R ^(j-n)×j ，

Step S45: calculating and expanding observable matrix gamma _i ：

Γ _i ＝U ₁ S ₁ ^1/2 Formula thirteen;

Γ _i-1 ＝first(i-1)l rows of Γ _i formula fourteen

Step S46: solving the linear system of equations yields matrices A, C and K:

wherein Y is _i|i ＝[y _i y _i+1 … y _i+j-1 ]∈R ^l×j ，

Is with Z _i And U _f Orthogonal noise terms;

step S47: identifying the model order by using MSVC criterion:

where T is the future output blockThe number of matrix elements of the Hankel matrix,

the square of matrix 2 norms formed by singular value identification values is represented by C (T), which is a function of T;

step S48: using least squares to pass through matrix A, C, Γ _i And Γ _i-1 The matrices B and D are solved.

2. The method for identifying the control model structure of the modularized direct-driven fan according to claim 1, which is characterized by comprising the following steps: the input of the active power control module in the first step is the active power reference value P of the fan _Wref Angular velocity omega of fan rotor _gen Dc bus voltage u _dc Output net side current value d-axis component i of current amplitude limiting link _gd The output of the module is the active power P output by the fan; the input of the reactive power control module is the DC bus voltage u _dc Output net side current value q axis component i with current amplitude limiting link _gq Reactive power reference value Q of fan _ref The output of the module is the reactive power Q output by the fan; the input of the current limiting module is a network side current value i _g And the active power P is output by the active power control module; the input of the reactive power limiting module is the active power P output by the active power control module and the reactive power Q output by the reactive power control module, and the output is the Q after limiting ₁ 。

3. The method for identifying the control model structure of the modularized direct-driven fan according to claim 2, which is characterized by comprising the following steps of: the third step comprises the following steps;

step A1, collecting an active power control model G ₁ Input signal u of (2) ₁ And output signal y ₁ And reactive power control model G ₂ Input signal u of (2) ₂ And output signal y ₂ ；

Step A2, using the input signal u ₁ Active power control model obtained by excitation identification

Obtain the output signal +.>

By input signal u ₂ Reactive power control model obtained by excitation identification>

Obtain the output signal +.>

Step A3, respectively y ₁ And (3) with

And y ₂ And->

Comparing, and if the fitting degree between the two is high, identifying to obtain a reliable model; if the fitting degree of the two is larger, the model obtained by identification is unreliable; judging by using a fitting ratio index FR, wherein the fitting ratio index is defined as:

where N is the number of samples acquired, the closer FR to 1 indicates the better fit, and otherwise indicates the worse fit.

Images