CN106125550A - The combined failure of a kind of high ferro traction rectifier device is estimated and fault tolerant control method - Google Patents

Abstract

The combined failure that the invention discloses a kind of high ferro traction rectifier device is estimated and fault tolerant control method, belong to fault diagnosis and fault-tolerant control technical field, so that higher-dimension rectifier fault model exists under variable element coupling uncertain condition, the commutator combined failure of High Speed Railway Trains is estimated and fault tolerant control method remains able to accurately estimate fault in Microsecond grade Diagnostic Time, and improves the function self-healing capability of system.The method comprises the steps: the first step: converted by combined failure equivalent space, sets up dimensionality reduction combined failure evolutionary model；Second step: use method of least square based on stable filter to carry out identification model parameter；3rd step: set up commutator closed loop Fault Estimation model according to the strategy such as Decoupling Analysis, feedforward compensation；4th step: setting up multivariable dynamic compensator to improve the robustness under rectifier fault state, the model reference adaptive controller that designs a model accordingly is to ensure that system has globally consistent Fast Convergent characteristic.

Description

The combined failure of a kind of high ferro traction rectifier device is estimated and fault tolerant control method

Technical field

The present invention relates to fault diagnosis and fault-tolerant control technical field, specifically, relate to a kind of High Speed Railway Trains and lead The combined failure drawing commutator is estimated and fault tolerant control method.

Background technology

Being different from the commutator of common three-phase alternating current electric drive, the employing of High Speed Railway Trains traction rectifier device is single-phase to supply in turn Power mode.When bullet train is along electric power phase-separating section long-play, this circulation driving mechanism is that commutator multivariate is multiple Close the root that fault occurs.In commutator four-quadrant Mode-switch running, although controller, executor and sensor are all It is normal, but inappropriate triggering pulse may still lead to the generation of combined failure.In addition, it is contemplated that contact net side and inverter High intensity bi-directional energy flow between the drive system of side, the complicated interleaved operation operating mode of high ferro traction rectifier device more susceptible to The impact of combined failure.The at present research about commutator mostly concentrates on the power protection of switch mode, high power quality and Frequency analyses etc., seldom consider the Fault Estimation under commutator variable element Coupled Dynamic and faults-tolerant control problem.

One key issue of multivariate Fault Tolerance Control Technology is that input and output are carried out reasonably " pairing " so that hand over The impact of interaction minimizes.As interactive a kind of useful metrics, relative gain array can ensure that multivariate is multiple Close the decoupling pairing of Fault Estimation.At present, high ferro rectifier fault diagnosis mechanism based on fault-safety principle, be a kind of towards The open loop method for diagnosing faults of traction control cell protection.This method for diagnosing faults afterwards is difficult to meet traction rectifier device and is combined The diagnosis requirement of fault Microsecond grade.Additionally, high ferro traction rectifier device variable element coupling ambiguous model is difficult to solve combined failure Forecasting problem, consequently, it is possible to cause the problem that commutator security performance is demoted.Therefore, closed loop fault based on parameter identification is estimated Meter method is significant to the health forecast of high ferro traction rectifier device this kind of variable element coupled system.

In order to reduce the adverse effect that combined failure brings, and improve the Synchronization Control performance of four-quadrant rectifier network, The double-direction model that traditional feedback control is difficult between reply multivariable control loops and Failure Observer does not mate to be estimated with fault Meter error.Single fault-tolerance approach based on feedback control, can't improve the requirement of systemic-function self-healing capability.

Summary of the invention

It is an object of the invention to, for the problems referred to above, the present invention discloses the combined failure of a kind of high ferro traction rectifier device and estimates Meter and fault tolerant control method, set up the multivariate evolutionary model of commutator combined failure；Use minimum based on stable filter Square law carrys out Dynamic Identification model parameter, and design feed-forward compensator decouples multivariate combined failure system accordingly；Proportional, integral Robust closed loop Failure Observer solves combined failure detection sensitivity and AF panel problem, can be effectively improved system Faults-tolerant control performance；By design Globally asymptotic convergence rule, model reference adaptive based on multivariable dynamic compensator Control method can be effectively improved the faults-tolerant control performance of system.

Realization the technical scheme is that, the present invention first passes through the conversion of combined failure equivalent space, by commutator four The mutual complex network Dynamic and Abstract of quadrant is the single failure parameter Evolution dynamics at four quadrants, sets up dimensionality reduction combined failure Evolutionary model；Propose variable element combined failure identification Method.The present invention builds according to the strategy such as Decoupling Analysis, feedforward compensation Vertical commutator closed loop Fault Estimation model, design Multi-objective Robust solver improves combined failure detection sensitivity accordingly.Root Estimate, with faults-tolerant control problem, there is duality feature according to combined failure, initially set up multivariable dynamic compensator to improve rectification Robustness under device malfunction, the model reference adaptive controller that designs a model accordingly is to ensure that system has globally consistent quick receipts Hold back characteristic.

The combined failure that the invention provides a kind of high ferro traction rectifier device is estimated and fault tolerant control method, the method step For:

(1) there is from traction rectifier device combined failure the principle of equivalent space, introduce the single failure evolution factor and come Describe commutator higher-dimension Evolution dynamics, set up and describe traction rectifier device combined failure characteristic, interference characteristic and the fall of structural damage Dimension module；Use method of least square based on stable filter to carry out Dynamic Identification model parameter, obtain the multivariate fault determined Model；According to the most recognized model parameter, the correlation analysis theoretical by relative gain array realizes multi-variable system Decoupling, design feed-forward compensator realizes the stability before closed loop system fault detect accordingly；It is combined for traction rectifier device The diagnosis of fault Microsecond grade and AF panel requirement, design multiple target proportional, integral closed loop robust Fault Estimation device.

(2), on the basis of the multivariate combined failure in above-mentioned foundation estimates model, coupling system is estimated and system controls Between duality principle, design Robust Dynamic Compensator improve the fault-tolerant ability under the worst failure condition of system, set accordingly Meter feedback of status model reference self-adapting control tackles the uncertainty of failure system.

In the present invention, traction rectifier device dimensionality reduction combined failure model is determined by following principle and method:

$\left\{\begin{array}{c}\stackrel{\·}{x}=A\left(\mathrm{\ξ}\right)x\left(t\right)+Bu\left(t\right)+E(t-t_f)f+D_{1}w\left(t\right)\\ y\left(t\right)=Cx\left(t\right)+D_{2}w\left(t\right)\end{array}\right.---\left(1\right)$

In formula, state variable x=[i_C V_dc] comprise current on line side i_cWith DC voltage V_dc；Jointly control input variable u =[i_ld V_g] it is defined as load current i_ldWith voltage on line side V_g；Control Loops For Multi-variable Control System is output as y=[i_C V_dc]；A (ξ) represents 2 × 2 Parametric System matrixes of dimensionality reduction, ξ (0≤ξ≤1) is the failure evolution factor；B is the input matrix of 2 × 2, and C is 2 × 2 Output matrix；F=[f₁ f₂] it is the voltage on line side and load current fault, t coupled_fFor unknown failure time of origin,Representative can recover the time profile of fault；W is the harmonic wave interference of energy bounded, D₁And D₂Respectively Represent perturbation matrix.

Due to A (ξ) and t in formula (1)_fIt is uncertain, traditional based on a determination that the Fault Estimation method of system is uncomfortable For the present invention program.In view of t_fUncertainty can be equivalent to the dynamic characteristic of ξ, first the present invention designs based on stable The method of least square of wave filter carrys out Dynamic Identification ξ.Therefore, formula (1) is reduced to following form:

$\left\{\begin{array}{c}\stackrel{\·}{x}\left(t\right)=A\left(\mathrm{\ξ}\right)x\left(t\right)+Bu\left(t\right)+{\mathrm{\Δ}}_1\\ y\left(t\right)=Cx\left(t\right)+{\mathrm{\Δ}}_2\end{array}\right.---\left(2\right)$

Δ in formula₁And Δ₂For acting on the unknown on sensor and executor but constant unfavorable conditions.

According to equivalence transformation principle, the s territory expression formula of formula (2) is:

P_ξ(s) [y] (t)=Z^Δ(s)[u](t)+π(s)[v](t) (3)

P in formula_ξ(s)=s³-a₂ξs²,Z^Δ(S)=b₂c₁s², π (s)=Δ₂s²+(c₁Δ₁-Δ₂a₂ξ) s is known multinomial Formula, v (t) is immeasurablel noise.

In order to estimate parameter value ξ accurately from noise signal v (t), select following Stable Polynomials

Λ (s)=s³+λ2₂s²+λ₁s+λ₀(all zero points are respectively positioned on s Left half-plane) carrys out the wave filter of reliable design Formula (3) both sides are filtered simultaneously, can obtain:

$y\left(t\right)=\frac{Z_{\mathrm{\Δ}}\left(s\right)}{\mathrm{\Λ}\left(s\right)}\[u\]\left(t\right)+\frac{\mathrm{\Λ}\left(s\right)-P_{\mathrm{\ξ}}\left(s\right)}{\mathrm{\Λ}\left(s\right)}\[y\]\left(t\right)+\frac{\mathrm{\π}\left(s\right)}{\mathrm{\Λ}\left(s\right)}\[\mathrm{\υ}\]\left(t\right)---\left(4\right)$

For the filtering of High-frequency Interference, can be as drawn a conclusion:

$\underset{s\→\mathrm{\∞}}{\lim }\frac{\mathrm{\π}\left(s\right)}{\mathrm{\Λ}\left(s\right)}\[\mathrm{\υ}\]\left(t\right)=\underset{s\→\mathrm{\∞}}{\lim }\frac{{\mathrm{\Δ}}_2{s}^2+(c_1{\mathrm{\Δ}}_1-{\mathrm{\Δ}}_2{a}_2\mathrm{\ξ})s}{s^3+{\mathrm{\λ}}_2{s}^2+{\mathrm{\λ}}_{1}s+{\mathrm{\λ}}_0}\[\mathrm{\υ}\]\left(t\right)=0---\left(5\right)$

Therefore, formula (4) can be expressed as:

$y\left(t\right)=\frac{Z_{\mathrm{\Δ}}\left(s\right)}{\mathrm{\Λ}\left(s\right)}\[u\]\left(t\right)+\frac{\mathrm{\Λ}\left(s\right)-P_{\mathrm{\ξ}}\left(s\right)}{\mathrm{\Λ}\left(s\right)}\[y\]\left(t\right)---\left(6\right)$

By introducing following parameter θ^*With regression vector φ (t):

θ^*=[b₂c₁ λ₀ λ₁ λ₂+a₂ξ]

$\mathrm{\φ}\left(t\right)=\[\begin{array}{cccc}\frac{s^2}{\mathrm{\Λ}\left(s\right)}\[u\]\left(t\right)& \frac{1}{\mathrm{\Λ}\left(s\right)}\[y\]\left(t\right)& \frac{s}{\mathrm{\Λ}\left(s\right)}\[y\]\left(t\right)& \frac{s^2}{\mathrm{\Λ}\left(s\right)}\[y\]\left(t\right)\end{array}\]$

Formula (6) can be converted into following parameterized form:

Y (t)=(θ^*)^Tφ(t) (7)

Therefore, standardized method of least square is used can to obtain unknown parameter θ in formula (7)^*Or the zero steady-state error of ξ estimates Meter.ξ identification obtained substitutes into formula (1), then uncertain parameter matrix A (ξ) converges to constant matrix A.In view of t_fThe most true The qualitative dynamic characteristic that can be equivalent to ξ, according to the ξ value of convergence, t_fZero steady-state error value also can estimate, i.e. in formula (1) Uncertain parameter matrix E (t-t_f) it is recognized as E.

In the present invention, multiple target proportional, integral closed loop robust Fault Estimation device is determined by following principle and method:

Be there is feedforward controller u (the t)=K of stabilization function by introducing₁Y (t), then the closed loop of formula (2) is:

$\stackrel{\·}{x}\left(t\right)=(A+{\mathrm{BK}}_{1}C+\mathrm{\Δ}\mathrm{\ξ})x\left(t\right)+{\mathrm{BK}}_1{\mathrm{\Δ}}_2+{\mathrm{\Δ}}_1---\left(8\right)$

In formulaFor parameter identification error.

According to formula (1) and formula (8), consider rapidity and the multiple target proportional, integral closed loop robust of capacity of resisting disturbance simultaneously Fault approximator can design as follows:

$\left\{\begin{array}{c}\frac{d\hat{x}}{dt}=(A+{\mathrm{BK}}_{1}C)\hat{x}\left(t\right)+E\hat{f}-L_p\left(\hat{y}\right(t)-y(t\left)\right)\\ \frac{d\hat{f}}{dt}=-L_I\left(\hat{y}\right(t)-y(t\left)\right)\\ \hat{y}\left(t\right)=C\hat{x}\left(t\right)\end{array}\right.---\left(9\right)$

L in formula_pAnd L_IIt is respectively proportional gain to be designed and storage gain,It is estimator state variable,Generation Table estimator output variable.

According to H_∞Robust Analysis and comprehensive theory, by selecting suitable positive definite symmetric matricesAnd matrixThen according to than Example-integration closed loop fault approximatorProportional gain L can be obtained_PWith storage gain L_I。

In the present invention, Robust Dynamic Compensator is determined by following principle and method:

On the basis of the combined failure set up in formula (9) estimates model, coupling system is estimated and system control between right Idol principle, the dynamic compensator of high ferro commutator may be designed as:

$\left\{\begin{array}{c}\stackrel{\·}{z}\left(t\right)=A_{K}z\left(t\right)+B_{K}y\left(t\right)\\ u\left(t\right)=C_{K}z\left(t\right)+D_{K}y\left(t\right)-E\hat{f}\end{array}\right.---\left(10\right)$

In formula, A_KFor the inverse transformation of high ferro commutator eigenmatrix, B_KCorrespond to the output matrix of commutator, C_KRepresentative is many Variable input matrix, D_KRepresent the feed-forward coefficients matrix of commutator.

Theoretical according to robust Fault-Tolerant Control, by selecting suitable positive definite symmetric matrices X, Y, four matrixWith Matrix M, N, the then parameter (A of dynamic compensator_K,B_K,C_K,D_K) can design as follows:

$\begin{array}{c}{D}_K=\hat{D}\\ C_K=(\hat{C}-D_{K}C\stackrel{\‾}{\mathrm{\ξ}}X)M^{-T}\\ B_K=N^{-1}(\hat{B}-{\mathrm{YBD}}_K)\\ A_K=N^{-1}(\hat{A}-\stackrel{\‾}{\mathrm{\ξ}}Y(A+{\mathrm{BD}}_{K}C\left)\stackrel{\‾}{\mathrm{\ξ}}X\right)M^{-T}\\ -B_{K}C\stackrel{\‾}{\mathrm{\ξ}}{\mathrm{XM}}^{-T}-N^{-1}\stackrel{\‾}{\mathrm{\ξ}}{\mathrm{YBC}}_K\end{array}---\left(11\right)$

In formula,M, N meet

In the present invention, feedback of status model reference self-adapting control is determined by following principle and method:

According to the method for designing of formula (11), the known portions in formula (10) and unknown portions can be separated and write, i.e.

$\stackrel{\·}{z}=(A_K-\frac{B_K{C}_K}{D_K})z+\frac{B_K}{D_K}\mathrm{\Λ}(u+{\mathrm{\ξ}}^{T}E)---\left(12\right)$

In formula, uncertain diagonal matrix Λ representative model controls fault effect.

By selecting following model reference self-adapting control rate

$\left\{\begin{array}{c}\frac{d{\hat{K}}_z}{dt}=-{\mathrm{\Γ}}_z{\mathrm{ze}}^T{P}_K\frac{B_K}{D_K}\\ \frac{d{\hat{K}}_r}{dt}=-{\mathrm{\Γ}}_{r}r\left(t\right)e^T{P}_K\frac{B_K}{D_K}\\ \frac{d\widehat{\mathrm{\ξ}}}{dt}={\mathrm{\Γ}}_{\mathrm{\ξ}}{\mathrm{Ee}}^T{P}_K\frac{B_K}{D_K}\end{array}\right.---\left(13\right)$

In formulaFor ideal feedback gain K_zEstimation,Represent preferable feedforward gain K_rEstimation,For the estimation of ξ, from Adaptation rate Γ_z, Γ_rAnd Г_ξIt is respectively known symmetric positive definite matrix.

Following condition is set up:

Reference model

A in formula_refFor Hurwitz matrix, r is bounded reference-input signal.

Model Matching condition

Tracking error e=z-z_ref (16)

Control input

Algebraically Lyapunov Equation

Then formula (12) achieves the globally consistent progressive tracking performance of reference model dynamic type (14).

Beneficial effect:

The present invention discloses the combined failure of a kind of high ferro traction rectifier device and estimates and fault tolerant control method, sets up commutator multiple Close the multivariate evolutionary model of fault；Method of least square based on stable filter is used to carry out Dynamic Identification model parameter, accordingly Design feed-forward compensator decouples multivariate combined failure system；Design proportion-integration robust closed loop Failure Observer solves Combined failure detection sensitivity and AF panel problem；By design Globally asymptotic convergence rule, mend based on multivariable dynamic The Model Reference Adaptive Control Method repaying device can be effectively improved the faults-tolerant control performance of system.

Accompanying drawing explanation

Fig. 1 is the equivalent space figure of combined failure；

Fig. 2 is that traction rectifier device combined failure is estimated and faults-tolerant control structure chart；

Fig. 3 is the Stable structure figure under system fault condition；

Fig. 4 is closed loop Fault Estimation schematic diagram；

Fig. 5 is combined failure evolution factor dynamic effect picture；

Fig. 6 is combined failure open loop estimation effect figure；

Fig. 7 is combined failure closed loop estimation effect figure；

Fig. 8 is system mode quality reconstruction figure；

Fig. 9 is system control variables variation effect figure.

Detailed description of the invention

Describe embodiments of the present invention in detail below with reference to drawings and Examples, whereby how the present invention is applied Technological means solves technical problem, and the process that realizes reaching technique effect can fully understand and implement according to this.Need explanation As long as not constituting conflict, each embodiment in the present invention and each feature in each embodiment can be combined with each other, The technical scheme formed is all within protection scope of the present invention.

The combined failure that the invention provides a kind of high ferro traction rectifier device is estimated and fault tolerant control method, the method step For:

(1) there is from traction rectifier device combined failure the principle of equivalent space, introduce the single failure evolution factor and come Describe commutator higher-dimension Evolution dynamics, set up and describe traction rectifier device combined failure characteristic, interference characteristic and the fall of structural damage Dimension module；Use method of least square based on stable filter to carry out Dynamic Identification model parameter, obtain the multivariate fault determined Model；According to the most recognized model parameter, the correlation analysis theoretical by relative gain array realizes multi-variable system Decoupling, design feed-forward compensator realizes the stability before closed loop system fault detect accordingly；It is combined for traction rectifier device The diagnosis of fault Microsecond grade and AF panel requirement, design multiple target proportional, integral closed loop robust Fault Estimation device.

(2), on the basis of the multivariate combined failure in above-mentioned foundation estimates model, coupling system is estimated and system controls Between duality principle, design Robust Dynamic Compensator improve the fault-tolerant ability under the worst failure condition of system, set accordingly Meter feedback of status model reference self-adapting control tackles the uncertainty of failure system.

In the present invention, traction rectifier device dimensionality reduction combined failure model is determined by following principle and method:

$\left\{\begin{array}{c}\stackrel{\·}{x}=A\left(\mathrm{\ξ}\right)x\left(t\right)+Bu\left(t\right)+E(t-t_f)f+D_{1}w\left(t\right)\\ y\left(t\right)=Cx\left(t\right)+D_{2}w\left(t\right)\end{array}\right.---\left(1\right)$

In formula, state variable x=[i_C V_dc] comprise current on line side i_cWith DC voltage V_dc；Jointly control input variable u =[i_ld V_g] it is defined as load current i_ldWith voltage on line side V_g；Control Loops For Multi-variable Control System is output as y=[i_C V_dc]；A (ξ) represents 2 × 2 Parametric System matrixes of dimensionality reduction, ξ (0≤ξ≤1) is the failure evolution factor；B is the input matrix of 2 × 2, and C is 2 × 2 Output matrix；F=[f₁ f₂] it is the voltage on line side and load current fault, t coupled_fFor unknown failure time of origin,Representative can recover the time profile of fault；W is the harmonic wave interference of energy bounded, D₁And D₂Respectively Represent perturbation matrix.

Due to A (ξ) and t in formula (1)_fIt is uncertain, traditional based on a determination that the Fault Estimation method of system is uncomfortable For the present invention program.In view of t_fUncertainty can be equivalent to the dynamic characteristic of ξ, first the present invention designs based on stable The method of least square of wave filter carrys out Dynamic Identification ξ.Therefore, formula (1) is reduced to following form:

$\left\{\begin{array}{c}\stackrel{\·}{x}\left(t\right)=A\left(\mathrm{\ξ}\right)x\left(t\right)+Bu\left(t\right)+{\mathrm{\Δ}}_1\\ y\left(t\right)=Cx\left(t\right)+{\mathrm{\Δ}}_2\end{array}\right.---\left(2\right)$

Δ in formula₁And Δ₂For acting on the unknown on sensor and executor but constant unfavorable conditions.

According to equivalence transformation principle, the s territory expression formula of formula (2) is:

P_ξ(s) [y] (t)=Z_Δ(s)[u](t)+π(s)[v](t) (3)

P in formula_ξ(s)=s³-a₂ξs²,Z^Δ(S)=b₂c₁s², π (s)=Δ₂s²+(c₁Δ₁-Δ₂a₂ξ) s is known multinomial Formula, υ (t) is immeasurablel noise.

In order to estimate parameter value ξ accurately from noise signal v (t), select following Stable Polynomials

Λ (s)=s³+λ2₂s²+λ₁s+λ₀(all zero points are respectively positioned on s Left half-plane) carrys out the wave filter of reliable designRight Formula (3) both sides are filtered simultaneously, can obtain:

$y\left(t\right)=\frac{Z_{\mathrm{\Δ}}\left(s\right)}{\mathrm{\Λ}\left(s\right)}\[u\]\left(t\right)+\frac{\mathrm{\Λ}\left(s\right)-P_{\mathrm{\ξ}}\left(s\right)}{\mathrm{\Λ}\left(s\right)}\[y\]\left(t\right)+\frac{\mathrm{\π}\left(s\right)}{\mathrm{\Λ}\left(s\right)}\[\mathrm{\υ}\]\left(t\right)---\left(4\right)$

For the filtering of High-frequency Interference, can be as drawn a conclusion:

$\underset{s\→\mathrm{\∞}}{\lim }\frac{\mathrm{\π}\left(s\right)}{\mathrm{\Λ}\left(s\right)}\[\mathrm{\υ}\]\left(t\right)=\underset{s\→\mathrm{\∞}}{\lim }\frac{{\mathrm{\Δ}}_2{s}^2+(c_1{\mathrm{\Δ}}_1-{\mathrm{\Δ}}_2{a}_2\mathrm{\ξ})s}{s^3+{\mathrm{\λ}}_2{s}^2+{\mathrm{\λ}}_{1}s+{\mathrm{\λ}}_0}\[\mathrm{\υ}\]\left(t\right)=0---\left(5\right)$

Therefore, formula (4) can be expressed as:

$y\left(t\right)=\frac{Z_{\mathrm{\Δ}}\left(s\right)}{\mathrm{\Λ}\left(s\right)}\[u\]\left(t\right)+\frac{\mathrm{\Λ}\left(s\right)-P_{\mathrm{\ξ}}\left(s\right)}{\mathrm{\Λ}\left(s\right)}\[y\]\left(t\right)---\left(6\right)$

By introducing following parameter θ^*With regression vector φ (t):

θ^*=[b₂c₁ λ₀ λ₁ λ₂+a₂ξ]

$\mathrm{\φ}\left(t\right)=\[\begin{array}{cccc}\frac{s^2}{\mathrm{\Λ}\left(s\right)}\[u\]\left(t\right)& \frac{1}{\mathrm{\Λ}\left(s\right)}\[y\]\left(t\right)& \frac{s}{\mathrm{\Λ}\left(s\right)}\[y\]\left(t\right)& \frac{s^2}{\mathrm{\Λ}\left(s\right)}\[y\]\left(t\right)\end{array}\]$

Formula (6) can be converted into following parameterized form:

Y (t)=(θ^*)^Tφ(t) (7)

Therefore, standardized method of least square is used can to obtain unknown parameter θ in formula (7)^*Or the zero steady-state error of ξ estimates Meter.ξ identification obtained substitutes into formula (1), then uncertain parameter matrix A (ξ) converges to constant matrix A.In view of t_fThe most true The qualitative dynamic characteristic that can be equivalent to ξ, according to the ξ value of convergence, t_fZero steady-state error value also can estimate, i.e. in formula (1) Uncertain parameter matrix E (t-t_f) it is recognized as E.

In the present invention, multiple target proportional, integral closed loop robust Fault Estimation device is determined by following principle and method:

Be there is feedforward controller u (the t)=K of stabilization function by introducing₁Y (t), then the closed loop of formula (2) is:

$\stackrel{\·}{x}\left(t\right)=(A+{\mathrm{BK}}_{1}C+\mathrm{\Δ}\mathrm{\ξ})x\left(t\right)+{\mathrm{BK}}_1{\mathrm{\Δ}}_2+{\mathrm{\Δ}}_1---\left(8\right)$

In formulaFor parameter identification error.

According to formula (1) and formula (8), consider rapidity and the multiple target proportional, integral closed loop robust of capacity of resisting disturbance simultaneously Fault approximator can design as follows:

$\left\{\begin{array}{c}\frac{d\hat{x}}{dt}=(A+{\mathrm{BK}}_{1}C)\hat{x}\left(t\right)+E\hat{f}-L_p\left(\hat{y}\right(t)-y(t\left)\right)\\ \frac{d\hat{f}}{dt}=-L_I\left(\hat{y}\right(t)-y(t\left)\right)\\ \hat{y}\left(t\right)=C\hat{x}\left(t\right)\end{array}\right.---\left(9\right)$

L in formula_pAnd L_IIt is respectively proportional gain to be designed and storage gain,It is estimator state variable,Generation Table estimator output variable.

According to H_∞Robust Analysis and comprehensive theory, by selecting suitable positive definite symmetric matricesAnd matrixThen according to than Example-integration closed loop fault approximatorProportional gain L can be obtained_PWith storage gain L_I。

In the present invention, Robust Dynamic Compensator is determined by following principle and method:

On the basis of the combined failure set up in formula (9) estimates model, coupling system is estimated and system control between right Idol principle, the dynamic compensator of high ferro commutator may be designed as:

$\left\{\begin{array}{c}\stackrel{\·}{z}\left(t\right)=A_{K}z\left(t\right)+B_{K}y\left(t\right)\\ u\left(t\right)=C_{K}z\left(t\right)+D_{K}y\left(t\right)-E\hat{f}\end{array}\right.---\left(10\right)$

In formula, A_KFor the inverse transformation of high ferro commutator eigenmatrix, B_KCorrespond to the output matrix of commutator, C_KRepresentative is many Variable input matrix, D_KRepresent the feed-forward coefficients matrix of commutator.

Theoretical according to robust Fault-Tolerant Control, by selecting suitable positive definite symmetric matrices X, Y, four matrixWith Matrix M, N, the then parameter (A of dynamic compensator_K,B_K,C_K,D_K) can design as follows:

$\begin{array}{c}{D}_K=\hat{D}\\ C_K=(\hat{C}-D_{K}C\stackrel{\‾}{\mathrm{\ξ}}X)M^{-T}\\ B_K=N^{-1}(\hat{B}-{\mathrm{YBD}}_K)\\ A_K=N^{-1}(\hat{A}-\stackrel{\‾}{\mathrm{\ξ}}Y(A+{\mathrm{BD}}_{K}C\left)\stackrel{\‾}{\mathrm{\ξ}}X\right)M^{-T}\\ -B_{K}C\stackrel{\‾}{\mathrm{\ξ}}{\mathrm{XM}}^{-T}-N^{-1}\stackrel{\‾}{\mathrm{\ξ}}{\mathrm{YBC}}_K\end{array}---\left(11\right)$

In formula,M, N meet

In the present invention, feedback of status model reference self-adapting control is determined by following principle and method:

According to the method for designing of formula (11), the known portions in formula (10) and unknown portions can be separated and write, i.e.

$\stackrel{\·}{z}=(A_K-\frac{B_K{C}_K}{D_K})z+\frac{B_K}{D_K}\mathrm{\Λ}(u+{\mathrm{\ξ}}^{T}E)---\left(12\right)$

In formula, uncertain diagonal matrix Λ representative model controls fault effect.

By selecting following model reference self-adapting control rate

$\left\{\begin{array}{c}\frac{d{\hat{K}}_z}{dt}=-{\mathrm{\Γ}}_z{\mathrm{ze}}^T{P}_K\frac{B_K}{D_K}\\ \frac{d{\hat{K}}_r}{dt}=-{\mathrm{\Γ}}_{r}r\left(t\right)e^T{P}_K\frac{B_K}{D_K}\\ \frac{d\widehat{\mathrm{\ξ}}}{dt}={\mathrm{\Γ}}_{\mathrm{\ξ}}{\mathrm{Ee}}^T{P}_K\frac{B_K}{D_K}\end{array}\right.---\left(13\right)$

In formulaFor ideal feedback gain K_zEstimation,Represent preferable feedforward gain K_rEstimation,For the estimation of ξ, Adaptive rate Γ_z, Γ_rAnd Γ_ξIt is respectively known symmetric positive definite matrix.

Following condition is set up:

Reference model

A in formula_refFor Hurwitz matrix, r is bounded reference-input signal.

Model Matching condition

Tracking error e=z-z_ref (16)

Control input

Algebraically Lyapunov Equation

Then formula (12) achieves the globally consistent progressive tracking performance of reference model dynamic type (14).

Embodiment 1

The present invention with the true model parameter of certain high ferro traction rectifier device experimental testing station as objective for implementation, traction rectifier device Sampling time be 100 microseconds, fault detect and fault time less than 300 microseconds.For variable element combined failure system, carry Going out a kind of closed loop Failure Observer method for designing, this Fault Estimation method is possible not only to accurately decouple multiple simultaneous event Barrier, additionally it is possible to meet the Microsecond grade fault diagnosis requirement of traction rectifier device；For the uncertainty of failure system, a kind of Shandong is proposed Rod adaptive fusion method, it is special that this faults-tolerant control scheme is possible not only to ensure that system mode has globally consistent asymptotic convergence Property, moreover it is possible to ensure that the change of control variable is less than permissible range.

Embodiment

As described in figures 1 and 5, a kind of high ferro traction rectifier device combined failure equivalent space model is embodiments provided Discrimination method:

Based on the combined failure evolution factor four quadrants of commutator identifier ξ (0.3111,0.3015,0.6889, 0.6985), the parameter value of formula (1) can be calculated as follows:

$A_1=\left[\begin{array}{cc}-7.5& -9.445\\ 343.4545& 0\end{array}\right],A_2=\left[\begin{array}{cc}-7.5& -9.925\\ 360.9091& 0\end{array}\right]$

$A_3=\left[\begin{array}{cc}-7.5& 9.445\\ -343.4545& 0\end{array}\right],A_4=\left[\begin{array}{cc}-7.5& 9.925\\ -360.9091& 0\end{array}\right]$

$B=\left[\begin{array}{cc}-25& 7.5\\ 272.7& -909.0909\end{array}\right],C=\left[\begin{array}{cc}1& 0\\ 0& 1\end{array}\right]$

$E=\left[\begin{array}{cc}-10& 3\\ -910& 273\end{array}\right],D_1={\left[\begin{array}{cc}1& 1\end{array}\right]}^T,D_2={\left[\begin{array}{cc}1& 1\end{array}\right]}^T$

From fig. 5, it can be seen that combined failure evolution factor ξ has stronger sensitivity in the fault transient stage, based on ladder The moment of degree change maximum can give the initial detection time that is out of order；Initial time based on steady-state process can give inspection of being out of order The upper limit time surveyed is 86 microseconds, can preferably meet the failure diagnosis time requirement of 300 microseconds.

As shown in Fig. 2,3,4 and 7, embodiments provide a kind of high ferro traction rectifier device combined failure closed loop and estimate Method:

(1) according to relative gain array theory analysis, high ferro traction rectifier device multivariate fault model should use non-diagonal Line pairing realizes decoupling, i.e. second control variable V_g(voltage on line side) controls first output variable i_C(current on line side), the One control variable i_ld(load current) regulates second output variable V_dc(DC voltage).This conclusion and practical engineering application It is consistent.

(2) by selecting suitable positive definite symmetric matricesCan design that to have quadratic form steady Determine the feedforward compensation gain of meaningEnsure the boundedness of each variable of failure system.

(3) for given norm H_∞Performance indications γ₁=0.6 and disc area Q₁(-0.64,0.64), is closed by selection Suitable positive definite symmetric matricesAnd matrixThen according to proportional, integral closed loop Fault approximatorProportional gain L can be obtained_PWith storage gain L_IIt is respectively as follows:

$L_P=\left[\begin{array}{c}-6.5\\ 884.0709\end{array}\right],L_I=\left[\begin{array}{c}-23.3707\\ -23.7544\end{array}\right]$

(4) from fig. 6, it can be seen that traditional combined failure open loop method of estimation can not pick out electric current and the direct current of coupling Current failure；And the fault-signal estimated still suffers from oscillatory occurences in the terminal stage of failure evolution.From figure 7 it can be seen that this The combined failure closed loop method of estimation that invention proposes can preferably realize electric current and the decoupling of DC voltage simultaneous faults；Estimate The final uniform convergence of fault-signal is to set-point, and has sinuso sine protractor, improves susceptiveness and the higher hamonic wave of Fault Estimation Inhibition.

As shown in FIG. 8 and 9, a kind of high ferro traction rectifier device combined failure faults-tolerant control side is embodiments provided Method:

(1) theoretical according to robust Fault-Tolerant Control, for given norm H_∞Performance indications γ₁=0.2 and disc area Q₁(- 0.535,0.535), by selecting suitable positive definite symmetric matrices X, Y, four matrixWith matrix M, N

$X=\left[\begin{array}{cc}29.4864& -0.7496\\ -0.7496& 39.7523\end{array}\right],Y=\left[\begin{array}{cc}5.134& 0.066\\ 0.066& 0.1201\end{array}\right]$

$\hat{A}=\left[\begin{array}{cc}-0.04& -0.0206\\ 11.6046& -0.1589\end{array}\right],\hat{B}=\left[\begin{array}{c}-50.2617\\ 17.7297\end{array}\right],\hat{C}=\left[\begin{array}{cc}33.9434& 0.7161\\ 26.6607& 36.6663\end{array}\right],\hat{D}=\left[\begin{array}{c}-0.0879\\ 0.9865\end{array}\right]$

$M=\left[\begin{array}{cc}-0.9999& 0.0121\\ 0.0121& 0.9999\end{array}\right],N=\left[\begin{array}{cc}101.3413& 2.4024\\ 0& 45.2481\end{array}\right]$

Then dynamic compensatorParameter (A_K,B_K,C_K,D_K) can design as follows:

$A_K=\left[\begin{array}{cc}49.0206& 34.5257\\ -0.3426& 5.7362\end{array}\right],B_K=\left[\begin{array}{cc}0.2762& -0.2886\\ -0.3582& 0.1578\end{array}\right],$

$C_K=\left[\begin{array}{cc}22.7016& -0.3678\\ 29.4575& -18.0681\end{array}\right],D_K=\left[\begin{array}{cc}0.1117& -0.0781\\ -0.5258& 0.6681\end{array}\right]$

(2) by selecting following adaptive rate Γ_z, Γ_rAnd Γ_ξ, it is ensured that the adaptive fusion of feedback of status Algorithm has globally consistent asymptotic convergence characteristic.

${\mathrm{\Γ}}_z=\left[\begin{array}{cc}0.99& 0\\ 0& 0.99\end{array}\right],{\mathrm{\Γ}}_r=0.0084$

${\mathrm{\Γ}}_{\mathrm{\ξ}}=\left[\begin{array}{cccc}0.0588& 0& 0& 0\\ 0& 0.0588& 0& 0\\ 0& 0& 0.0588& 0\\ 0& 0& 0& 0.0588\end{array}\right]$

(3) from Fig. 8 and Fig. 9 it can be seen that under the system condition of uncertainty such as the system failure and higher harmonic current interference, The state of system reconfiguration and control variable still have preferable Decoupling Characteristics and zero steady-state error tracking effect.It addition, with independence Closed loop Fault Estimation compare (Fig. 7), the method (Fig. 8) of Fault Estimation and faults-tolerant control Integrated design has preferably interference and presses down Effect processed, higher decoupling and function self-healing capability.

Claims (8)

1. the combined failure of high ferro traction rectifier device is estimated and a fault tolerant control method, it is characterized in that, sets up rectifier systems The multivariate evolutionary model of combined failure, uses method of least square based on stable filter to come Dynamic Identification model parameter, root According to described model parameter design feed-forward compensator decoupling multivariate combined failure model；Design proportion-integration robust closed loop fault Observer solves combined failure detection sensitivity and AF panel problem, improves the faults-tolerant control performance of system；By design Globally asymptotic convergence rule, model reference self-adapting control based on multivariable dynamic compensator improves the faults-tolerant control of system Performance.

High ferro traction rectifier device combined failure the most according to claim 1 is estimated and fault tolerant control method, it is characterized in that, institute The method of stating comprises the steps of

Step one, has the principle of equivalent space from traction rectifier device combined failure, introduces the single failure evolution factor and comes Describe rectifier systems higher-dimension Evolution dynamics, set up and describe traction rectifier device system combined failure characteristic, interference characteristic and structure The dimensionality reduction model of damage；

Step 2, uses method of least square based on stable filter to carry out model parameter described in Dynamic Identification, obtains determining many Variable fault model；And according to the most recognized model parameter, decouple described multivariate by relative gain array theory and be combined Fault model, design feed-forward compensator realizes the stability before closed loop system fault detect；

Step 3, designs multiple target proportional, integral closed loop robust Fault Estimation device, it is achieved traction rectifier device combined failure Microsecond grade Diagnosis and to AF panel；

Step 4, coupling system estimates that the duality principle between system control, design Robust Dynamic Compensator improve system System fault-tolerant ability, design point feedback model Model Reference Adaptive Control tackles the uncertainty of failure system.

High ferro traction rectifier device combined failure the most according to claim 2 is estimated and fault tolerant control method, it is characterized in that, institute The multivariate evolutionary model stating multivariate combined failure is:

$\left\{\begin{array}{c}\stackrel{\·}{x}=A\left(\mathrm{\ξ}\right)x\left(t\right)+Bu\left(t\right)+E(t-t_f)f+D_{1}w\left(t\right)\\ y\left(t\right)=Cx\left(t\right)+D_{2}w\left(t\right)\end{array}\right.---\left(1\right)$

In formula, state variable x=[i_C V_dc] for comprising current on line side i_cWith DC voltage V_dc；Jointly control input variable u= [i_ld V_g] it is defined as load current i_ldWith voltage on line side V_g；Control Loops For Multi-variable Control System is output as y=[i_C V_dc]；A (ξ) represents fall The Parametric System matrix of dimension, ξ (0≤ξ≤1) is the failure evolution factor；B is input matrix, and C is output matrix；F=[f₁ f₂] Voltage on line side and load current fault, t for coupling_fFor unknown failure time of origin,Represent The time profile of fault can be recovered；W is the harmonic wave interference of energy bounded, D₁And D₂Represent perturbation matrix respectively.

High ferro traction rectifier device combined failure the most according to claim 3 is estimated and fault tolerant control method, it is characterized in that, base Method of least square in stable filter carrys out Dynamic Identification failure evolution factor ξ, and formula (1) is reduced to following form:

Δ in formula₁And Δ₂For acting on the unknown on sensor and executor but constant unfavorable conditions.

High ferro traction rectifier device combined failure the most according to claim 4 is estimated and fault tolerant control method, it is characterized in that,

According to equivalence transformation principle, the s territory expression formula of formula (2) is:

P_ξ(s) [y] (t)=Z_Δ(s)[u](t)+π(s)[υ](t) (3)

P in formula_ξ(s)=s³-a₂ξs²,Z_Δ(s)=b₂c₁s², π (s)=Δ₂s²+(c₁Δ₁-Δ₂a₂ξ) s is known multinomial, υ T () is immeasurablel noise；

In order to estimate parameter value ξ accurately from noise signal υ (t), select following Stable Polynomials

Λ (s)=s³+λ₂s²+λ₁sλ₀, all zero points are respectively positioned on s Left half-plane and carry out the wave filter of reliable designTo formula (3) Both sides are filtered simultaneously, can be as drawn a conclusion:

$\underset{s\→\mathrm{\∞}}{\lim }\frac{\mathrm{\π}\left(s\right)}{\mathrm{\Λ}\left(s\right)}\[\mathrm{\υ}\]\left(t\right)=\underset{s\→\mathrm{\∞}}{\lim }\frac{{\mathrm{\Δ}}_2{s}^2+(c_1{\mathrm{\Δ}}_1-{\mathrm{\Δ}}_2{a}_2\mathrm{\ξ})s}{s^3+{\mathrm{\λ}}_2{s}^2+{\mathrm{\λ}}_{1}s+{\mathrm{\λ}}_0}\[\mathrm{\υ}\]\left(t\right)=0---\left(5\right)$

$y\left(t\right)=\frac{Z_{\mathrm{\Δ}}\left(s\right)}{\mathrm{\Λ}\left(s\right)}\[u\]\left(t\right)+\frac{\mathrm{\Λ}\left(s\right)-P_{\mathrm{\ξ}}\left(s\right)}{\mathrm{\Λ}\left(s\right)}\[y\]\left(t\right)---\left(6\right)$

By introducing following parameter θ^*With regression vector φ (t):

θ^*=[b₂c₁ λ₀ λ₁ λ₂+a₂ξ]

$\mathrm{\φ}\left(t\right)=\left[\begin{array}{cccc}\frac{s^2}{\mathrm{\Λ}\left(s\right)}\[u\]\left(t\right)& \frac{1}{\mathrm{\Λ}\left(s\right)}\[y\]\left(t\right)& \frac{s}{\mathrm{\Λ}\left(s\right)}\[y\]\left(t\right)& \frac{s^2}{\mathrm{\Λ}\left(s\right)}\[y\]\left(t\right)\end{array}\right]$

Formula (6) can be converted into following parameterized form:

Y (t)=(θ^*)^Tφ(t) (7)

Standardized method of least square is used to obtain unknown parameter θ in formula (7)^*Or the zero steady-state error estimation of ξ.

High ferro traction rectifier device combined failure the most according to claim 5 is estimated and fault tolerant control method, it is characterized in that, adopts It is recognized as A by the uncertain parameter matrix A (ξ) in same procedure formula (1), and estimates to obtain t_fZero steady-state error value, i.e. Uncertain parameter matrix E (t-t in formula (1)_f) be recognized as E, by introducing have feedforward controller u (t) of stabilization function= K₁Y (t), then the closed loop of formula (2) is:

$\stackrel{\·}{x}\left(t\right)=(A+{\mathrm{BK}}_{1}C+\mathrm{\Δ}\mathrm{\ξ})x\left(t\right)+{\mathrm{BK}}_1{\mathrm{\Δ}}_2+{\mathrm{\Δ}}_1---\left(8\right)$

In formulaFor parameter identification error；

According to formula (1) and formula (8), consider rapidity and the multiple target proportional, integral closed loop robust Fault of capacity of resisting disturbance simultaneously Estimator design is as follows:

$\left\{\begin{array}{c}\frac{d\hat{x}}{dt}=\left(A+{\mathrm{BK}}_{1}C\right)\hat{x}\left(t\right)+E\hat{f}-L_p\left(\hat{y}\left(t\right)-y\left(t\right)\right)\\ \frac{d\hat{f}}{dt}=-L_I\left(\hat{y}\left(t\right)-y\left(t\right)\right)\\ \hat{y}\left(t\right)=C\hat{x}\left(t\right)\end{array}\right.---\left(9\right)$

L in formula_pAnd L_IIt is respectively proportional gain to be designed and storage gain,It is estimator state variable,Representative is estimated Gauge output variable；

According to H_∞Robust Analysis and comprehensive theory, by selecting suitable positive definite symmetric matricesAnd matrixThen according to ratio- Integration closed loop fault approximatorProportional gain L can be obtained_PWith storage gain L_I。

High ferro traction rectifier device combined failure the most according to claim 6 is estimated and fault tolerant control method, it is characterized in that, On the basis of the combined failure that formula (9) is set up estimates model, coupling system estimates the duality principle between system control, high The dynamic compensator of ferrum commutator may be designed as:

$\left\{\begin{array}{c}\stackrel{\·}{z}\left(t\right)=A_{K}z\left(t\right)+B_{K}y\left(t\right)\\ u\left(t\right)=C_{K}z\left(t\right)+D_{K}y\left(t\right)-E\hat{f}\end{array}\right.---\left(10\right)$

In formula, A_KFor the inverse transformation of high ferro commutator eigenmatrix, B_KCorrespond to the output matrix of commutator, C_KRepresent multivariate Input matrix, D_KRepresent the feed-forward coefficients matrix of commutator；

Theoretical according to robust Fault-Tolerant Control, by selecting suitable positive definite symmetric matrices X, Y, four matrixAnd matrix M, N, the then parameter (A of dynamic compensator_K,B_K,C_K,D_K) design as follows:

$\begin{array}{c}{D}_K=\hat{D}\\ C_K=(\hat{C}-D_{K}C\stackrel{\‾}{\mathrm{\ξ}}X)M^{-T}\\ B_K=N^{-1}\left(\hat{B}-{\mathrm{YBD}}_K\right)\\ A_K=N^{-1}\left(\hat{A}-\stackrel{\‾}{\mathrm{\ξ}}Y\left(A+{\mathrm{BD}}_{K}C\right)\stackrel{\‾}{\mathrm{\ξ}}X\right)M^{-T}\\ -B_K\stackrel{\‾}{C}\stackrel{\‾}{\mathrm{\ξ}}{\mathrm{XM}}^{-T}-N^{-1}\stackrel{\‾}{\mathrm{\ξ}}{\mathrm{YBC}}_K\end{array}---\left(11\right)$

In formula,M, N meet

High ferro traction rectifier device combined failure the most according to claim 7 is estimated and fault tolerant control method, it is characterized in that, root According to the method for designing of formula (11), the known portions in formula (10) and unknown portions are separated and write, i.e.

$\stackrel{\·}{z}=\left(A_K-\frac{B_K{C}_K}{D_K}\right)z+\frac{B_K}{D_K}\mathrm{\Λ}(u+{\mathrm{\ξ}}^{T}E)---\left(12\right)$

In formula, uncertain diagonal matrix Λ representative model controls fault effect；

By selecting following model reference self-adapting control rate

$\left\{\begin{array}{c}\frac{d{\hat{K}}_z}{dt}=-{\mathrm{\Γ}}_z{\mathrm{ze}}^T{P}_K\frac{B_K}{D_K}\\ \frac{d{\hat{K}}_r}{dt}=-{\mathrm{\Γ}}_{r}r\left(t\right)e^T{P}_K\frac{B_K}{D_K}\\ \frac{d\mathrm{\ξ}}{dt}={\mathrm{\Γ}}_{\mathrm{\ξ}}{\mathrm{Ee}}^T{P}_K\frac{B_K}{D_K}\end{array}\right.---\left(13\right)$

In formulaFor ideal feedback gain K_zEstimation,Represent preferable feedforward gain K_rEstimation,For the estimation of ξ, self adaptation Rate Γ_z, Γ_rAnd Γ_ξIt is respectively known symmetric positive definite matrix；

Following condition is set up:

Reference model

A in formula_refFor Hurwitz matrix, r is bounded reference-input signal；

Model Matching condition

Tracking error e=z-z_ref (16)

Control input

Algebraically Lyapunov Equation

Then formula (12) achieves the globally consistent progressive tracking performance of reference model dynamic type (14).