CN106014877A - Multi-fault diagnosis and fault-tolerant control of wind turbine system - Google Patents

Abstract

The invention discloses a multi-fault diagnosis and fault-tolerant control method of a wind turbine system. The method comprises the following steps: firstly, establishing a global fuzzy model of the wind turbine system by utilizing the T-S fuzzy algorithm; converting an actuator fault into a sensor fault by utilizing the characteristic of combining the sensor hardware redundancy technology with the actuator fault, and establishing a multi-fault diagnosis logical table to realize multi-fault detection; then, introducing a filter, converting the sensor fault into the actuator fault, establishing a virtual actuator fault, and realizing simultaneous reconstitution of two faults through reconstitution of the virtual actuator fault; finally, revising input and output of a controller based on a fault reconstitution value to realize active fault-tolerant control. The method has the following advantages: diagnosis and reconstitution of the wind turbine system with concurrent actuator and sensor faults can be simultaneously realized, accurate fault information is obtained in a real-time and on-line manner, fault-tolerant control is realized, the capacity of processing unknown faults of the system is enhanced, and the wind energy conversion efficiency under the fault is improved.

Description

Wind generator system multi-fault Diagnosis and faults-tolerant control

Technical field

The present invention relates to multi-fault Diagnosis and the faults-tolerant control of a kind of wind generator system, the most real-time based on FDD unit The accurate fault message obtained carries out active tolerant control to fault.

Background technology

The wind-powered electricity generation energy is regenerative resource the most with fastest developing speed, has become as solution energy issue of world indispensable Important force.Wind power plant is normally at high mountain or the remote field away from seashore, and climate change is unpredictable, at this In the working environment that sample height is severe, complicated, executor, sensor fault occur frequently.

Wind generator system internal structure is complicated, and component is numerous, and wherein actuator is the most complicated, including power train System, electricity generation system, yaw system and hydraulic system etc., the load of carrying is maximum, the maximum probability broken down, the peace to system Full reliability effect is the most maximum.Being safeguards system reliability service, internal system needs to install multiple sensor, as far as possible will simultaneously Comprehensively data are sent to data acquisition module, and by data acquisition module, these data are sent to master control system, through undue After analysis and process, send relevant control instruction, if these sensor failure, the operation of system can be directly influenced, especially , if the output signal of fault sensor is fed back in system controller, systematic function will be caused to reduce the most out of control.For this Wind generator system is carried out real-time fault diagnosis, implements effective faults-tolerant control to guarantee the reliability gesture of wind generator system Must go.

Fault diagnosis is as one of the important support technology of active tolerant control, and the system that enhances processes the energy of unknown failure Power, but can only judge whether system breaks down, failure reconfiguration supplements as the strong of fault diagnosis, realize fault detect and While separation, can get the information such as fault size and time of origin, for taking effective faults-tolerant control measure to eliminate fault to being The impact of system provides foundation more fully.

The diagnosis that research at present is both under wind generator system single-sensor or actuator failures mostly is with fault-tolerant Controlling, along with improving constantly of automatization level, wind generator system increasingly tends to complicate, and the situation that fault occurs is also multiple Miscellaneous changeable, in real system, single-sensor or actuator failures probability of happening are relatively low, and research has the biggest limitation.Wind simultaneously Power power-generating control system relies heavily on executor, sensor, and data-query interfaces and guarantees controlled system and control dress Have the most mutual between putting.Thus actuator and sensor failures is concurrently present in system often, to multiple faults Diagnosis and faults-tolerant control have more practical value.Therefore herein on the basis of multiple-fault classifier, to actuator failures and sensor Fault reconstructs simultaneously, provides sufficient real time fail information for system, reduces system failure rate by Fault Tolerance Control Technology, Reduce the maintenance of equipment time, it is ensured that system remains able to stable operation in certain indication range under fault, improve under fault Wind energy utilization efficiency.

Summary of the invention

It is an object of the invention to, for the executor in wind generator system and the simultaneous situation of sensor, design one Kind of multi-fault Diagnosis strategy and fault tolerant control method are for processing the wind generator system under multiple faults, it is ensured that system fault without When method obtains on-call maintenance, system also can maintain satisfied performance indications, it is achieved wind energy maximum capture under fault.

The technical method provided according to invention, described wind generator system multi-fault Diagnosis includes following with faults-tolerant control Step:

The first step: set up the state equation of wind generator system

The power generation process of double-fed wind-energy changing system is that wind energy is converted to mechanical energy, then is converted to electric energy by mechanical energy Process, in this process, the process of wind turbine capture electric energy is crucial, and it directly determines the conversion effect of wind-energy changing system Rate.

The most theoretical according to shellfish, mechanical output, pneumatic torque power coefficient and tip speed ratio that wind turbine produces are expressed Formula is

$P_{wt}=\frac{1}{2}{\mathrm{\ρ\πR}}^2{v}^3{C}_p(\mathrm{\λ},\mathrm{\β})---\left(1\right)$

${\mathrm{\Γ}}_{wt}=\frac{1}{2}{\mathrm{\ρ\πR}}^3{v}^2{C}_{\mathrm{\Γ}}(\mathrm{\λ},\mathrm{\β})---\left(2\right)$

C_p(λ, β)=λ C_Γ(λ,β) (3)

$\mathrm{\λ}=\frac{{\mathrm{R\Ω}}_l}{v}---\left(4\right)$

In formula, P_wtFor the capture power of wind energy conversion system, Γ_wtTorque, C is exported for wind mill wind wheel_Γ(λ, β) is wind energy conversion system Moment coefficient, C_P(λ, β) is power coefficient, and λ is the tip speed ratio of wind energy conversion system, and β is propeller pitch angle, and R is that pneumatic equipment blades is long Degree, ρ is atmospheric density, and v is wind speed, Ω_lFor wind wheel angular velocity.

Ignore some dynamic characteristics in model, owing to electromagnetic time constant is less than mechanical time constant, at system modelling Time negligible electromotor the dynamic process of electromagnetic response, then wind-energy changing system dynamical state equation can be expressed as follows:

$\left\{\begin{array}{c}\stackrel{\·}{x}\left(t\right)=A^{\′}x\left(t\right)+Bu\left(t\right)\\ y\left(t\right)=Cx\left(t\right)\end{array}\right.---\left(5\right)$

In formula, C=[10],x(t)∈RⁿFor state Vector, u (t) ∈ R^mFor input vector, y (t) ∈ R^pFor output vector, A' ∈ R^n×n, B ∈ R^n×n, C ∈ R^n×nIt is respectively system square Battle array, Ω_hFor the rotating speed of high speed shaft, Γ_GFor the electromagnetic torque of electromotor,For electromagnetic torque reference value, i is the speed change of gear Ratio, T_GFor electromagnetic time constant, J_tInertia is turned for high speed shaft.

Time below rated wind speed, pitch is fixed, it may be assumed that

${\mathrm{\Γ}}_{wt}({\mathrm{\Ω}}_h/i,v)=0.5{\mathrm{\ρ\πR}}^5{C}_p\left(\mathrm{\λ}\right){\mathrm{\Ω}}_h^2---\left(6\right)$

The control input understanding system from the state equation (5) of wind-energy changing system is electromagnetic torque reference value, and system Matrix A ' middle elements A₁₁It is as state variable Ω_hChange, so wind-energy changing system has strong nonlinear characteristic.

Second step: set up wind generator system overall situation T-S fuzzy model

Wind generator system carrying out T-S obscure, definition former piece fuzzy variable is z₁(t)=v, z₂(t)=Ω_h(t), then In system (1), matrix A can be write as new model A (z₁(t),z₂(t)).Take v₁≤min(v)≤max(v)≤v_m, Ω_h1≤min (Ω_h)≤max(Ω_h)≤Ω_hm.At interval [v₁,v_mM-2 point is taken on], and at interval [Ω_h1,Ω_hmAlso n-2 is taken on] Point, then available two sequences being made up of several former piece fuzzy variable values:

$\left\{\begin{array}{c}{Z}_1=(v_1,v_2,...,v_i,...v_m)\\ Z_2=({\mathrm{\Ω}}_{h1},{\mathrm{\Ω}}_{h2},...{\mathrm{\Ω}}_{hi},...{\mathrm{\Ω}}_{hn})\end{array}\right.---\left(7\right)$

In formula, i=1,2 ..., m, j=1,2 ..., n.

According to Taylor's formula, obtain wind torque variable at (v_i,Ω_hi) linear representation at place is:

${\mathrm{d\Γ}}_{wt}=\frac{{\mathrm{\γ\π\ρR}}^3}{2}{C}_{\mathrm{\Γ}}\left({\mathrm{\λ}}_i\right)\frac{i_0^2{v}_i^2}{{\mathrm{\Ω}}_{hi}}{\mathrm{d\Ω}}_h+\frac{(2-\mathrm{\γ}){\mathrm{\π\ρR}}^3}{2}{C}_{\mathrm{\Γ}}\left({\mathrm{\λ}}_i\right)v_{i}dv---\left(8\right)$

In formula,For torque parameter,

Formula (8) is substituted into formula (5), and by sequence Z₁And Z₂Middle element matches each other, and replaces sytem matrix A (z₁(t),z₂ (t)) in z₁(t) and z₂T (), can obtain a series of constant value matrix A_ij, i=1,2 ..., m, j=1,2 ..., n, further may be used The Local Linear Model obtaining i-th subsystem is:

$\stackrel{\·}{x}\left(t\right)=A_{i}x\left(t\right)+B_{i}u\left(t\right),i=1,2,...,r---\left(9\right)$

The fuzzy rule of definition T-S fuzzy model is as follows:

Rule Rⁱ:

If z₁(t)is M_i1andz₂is M_i2and…z_r(t)is M_ir (10)

$Then\left\{\begin{array}{c}\stackrel{\·}{x}\left(t\right)=A_{i}x\left(t\right)+B_{i}u\left(t\right)\\ y\left(t\right)=C_{i}x\left(t\right)+F{f}_s\end{array}\right.$

In formula, RⁱIt is i-th fuzzy rule, z_j(j=1,2) is former piece fuzzy variable, M_ijFor fuzzy subset, i=1, 2 ..., r, r are fuzzy rule sum.

Ambiguity in definition weights are:

$h_i\left(z\right(t\left)\right)=\frac{\underset{j=1}{\overset{n}{\mathrm{\Π}}}{M}_{ij}\left(z_j\right(t\left)\right)}{\underset{i=1}{\overset{r}{\mathrm{\Π}}}\underset{j=1}{\overset{n}{\mathrm{\Π}}}{M}_{ij}\left(z_j\right(t\left)\right)}---\left(11\right)$

In formula, i=1,2 ..., r, M_ij(z_j(t)) it is former piece fuzzy variable z_jT () be correspondence under i-th fuzzy rule Membership function, and the membership function of two fuzzy subsets all uses trigonometric function, and0≤h_i(z)≤1。

Then the global state equation of the wind generator system T-S fuzzy model under executor and sensor is:

$\left\{\begin{array}{c}\stackrel{\·}{x}\left(t\right)=\stackrel{\‾}{A}x\left(t\right)+\stackrel{\‾}{B}u\left(t\right)+{\mathrm{Df}}_a\left(t\right)\\ y\left(t\right)=\stackrel{\‾}{C}x\left(t\right)+{\mathrm{Ff}}_s\left(t\right)\end{array}\right.---\left(12\right)$

In formula, f_aRepresenting actuator failures, D is actuator failures Allocation matrix, f_sRepresenting sensor fault, F is sensor fault allocation matrix.

3rd step: wind generator system multi-fault Diagnosis strategy

Sensor fault in wind generator system is as a example by high speed shaft speed probe, and employing hardware redundancy technology is being System output high speed shaft rotating speed Ω_hThe position of signal is installed two duplicate speed probes and is used for measuring high speed shaft rotating speed, It is designated as Ω_h,m1And Ω_h,m2, i=1,2.When system is without sensor fault, Ω_h,m1=Ω_h,m2=Ω_h；When system generation sensor During fault, this signal becomes Ω_h,m1=Ω_h+ΔΩ_h,m1, Ω_h,_m2=Ω_h+ΔΩ_h,m2.Wherein, additional Δ Ω_h,m1And Δ Ω_h,m2Represent speed probe 1 and the contingent fault value of speed probe 2.Actuator failures in wind generator system May mainly have in source: the partial failures such as tooth surface abrasion, fatigue and the bearing deformation in drive system middle gear case, bearing；Oar In system fluid reveal excessive, oily in air content is higher and stop valve inefficacy etc., these faults can be modeled as parameter Deviation.

Wind generator system fault diagnosis: can be by comparing the measurement between two same type of sensors of same position Value, realizes fault diagnosis purpose.Difference between these two signals can directly show the generation of fault, such as, if | | Ω_h,m1 -Ω_h,m2| | in certain time period, it is not equal to 0, then system jam is described.

After fault detect completes, it is judged that system jam, further work find exactly which sensor or Executor breaks down.High speed shaft speed estimate can be accurately drawn by the T-S fuzzy-sliding-mode observer designed by a upper chapter ValueIt is the most closely actual value before and after any sensor fault occurs.Owing to actuator failures is internal system fault, It can cause system relevant parameter to change, and eventually affects the output of system, and therefore it can be considered sensor fault, and Can cause there is deviation between the measured value of two sensors and system normal output values.NoteWith Make residual signals, then system fault diagnosis strategy specifically describes, as shown in table 1:

Table 1 multi-fault Diagnosis logical table

4th step: actuator and sensor failures reconstructs simultaneously

Due to the increasing substantially of blower fan automatization level, internal system structure is complicated, fault type is numerous, and wind-force Electricity generation system relies heavily on executor, sensing and data query/interface to guarantee controlled system and to control there is conjunction between device Suitable is mutual, thus sensor and actuator fault reconstructs simultaneously and more has practical value and significance

Consider sensor fault and the simultaneous system of actuator failures (12), it is assumed that system (12) meets i.e.And the constant zero point of system is stable.Carry out obtaining such as down conversion to system (12):

Use orthogonal matrix T_RIt is multiplied by the system output matrix in formula (12), output equation can be decomposed into and not contain sensor Fault (y₁) and containing sensor fault (y₂) two measurement equations:

$y_1\left(t\right)={\stackrel{\‾}{C}}_{1}x\left(t\right)---\left(13\right)$

$y_2\left(t\right)={\stackrel{\‾}{C}}_{2}x\left(t\right)+F_1{f}_s\left(t\right)---\left(14\right)$

In order to realize that two kinds of faults are reconstructed simultaneously, following conversion thinking, an effective method is used just to be introduced into Wave filter, defines one group of new state variable z (t) as the output y containing sensor fault₂T () filtered signal, such as formula (15):

$\stackrel{\·}{z}\left(t\right)=-A_{f}z\left(t\right)+A_f{y}_2\left(t\right)=-A_{f}z\left(t\right)+A_f{\stackrel{\‾}{C}}_{2}x\left(t\right)+A_f{F}_1{f}_s\left(t\right)---\left(15\right)$

In formula, A_fIt it is a stationary filtering matrix.

Simultaneous formula (12) can obtain to (15)

$\left[\begin{array}{c}\stackrel{\·}{x}\left(t\right)\\ \stackrel{\·}{z}\left(t\right)\end{array}\right]=\left[\begin{array}{cc}\stackrel{\‾}{A}& 0\\ A_f{\stackrel{\‾}{C}}_2& -A_f\end{array}\right]\left[\begin{array}{c}x\left(t\right)\\ z\left(t\right)\end{array}\right]+\left[\begin{array}{c}\stackrel{\‾}{B}\\ 0\end{array}\right]u\left(t\right)+\left[\begin{array}{cc}D& 0\\ 0& A_f{F}_1\end{array}\right]\left[\begin{array}{c}{f}_a\left(t\right)\\ f_s\left(t\right)\end{array}\right]---\left(16\right)$

$\left[\begin{array}{c}{y}_1\left(t\right)\\ z\left(t\right)\end{array}\right]=\left[\begin{array}{cc}{\stackrel{\‾}{C}}_1& 0\\ 0& I_h\end{array}\right]\left[\begin{array}{c}x\left(t\right)\\ z\left(t\right)\end{array}\right]---\left(17\right)$

Consideration formula (16), (17), then wind generator system can be expressed as again:

${\stackrel{\·}{x}}_a\left(t\right)=A_a{x}_a\left(t\right)+B_{a}u\left(t\right)+{\mathrm{Mf}}_{as}\left(t\right)---\left(18\right)$

y_a(t)=C_ax_a(t) (19)

In formula, Represent virtual holding Row device fault vectors, it is also norm-bounded, i.e. f_as(t)≤η(x,u,t)+η₀, η₀It is a normal number, x_a(t)=[x (t )^Τ z(t)^Τ]^ΤFor the system state variables after coordinate transform.

By the above-mentioned conversion to system (12), it can be seen that new system (18) after conversion, (19) have not contained biography Sensor fault, but sensor fault is all become with actuator failures actuator failures and processes.The most only need to design phase It is theoretical that the sliding formwork Failure Observer answered combines output injection of equal value, can obtain virtual faults reconstruction valueSo far two kinds of faults f_s (t)、f_aT while (), reconstruction is addressed.

The design process of brief description wind generator system active tolerant control device:

The existence condition of T-S fuzzy-sliding-mode observer:

(1)rank(C_aM)=rank (M)；

(2)(A_a,M,C_a) constant zero point be stable.

The form of design wind generator system sliding mode observer is as follows:

${\stackrel{\·}{\hat{x}}}_a\left(t\right)=A_a{\hat{x}}_a+B_{a}u-G_l{e}_y+G_n{v}_{\mathrm{\ρ}}---\left(20\right)$

${\hat{y}}_a=C_a{\hat{x}}_a---\left(21\right)$

In formula, ρ is correction factor (gain), fault amplitude determined.G_l, G_nAlignment is treated for Robust Sliding Mode Observer Property feedback matrix and nonlinear feedback matrix.In order to guarantee sliding formwork motion and design sliding mode control strategy v_ρ, i.e.

$v_{\mathrm{\ρ}}=\{\begin{array}{cc}-\mathrm{\ρ}\left|\right|M_{2L}\left|\right|\frac{P_2{e}_{ya}}{\left|\right|P_2{e}_{ya}\left|\right|}& e_{ya}\≠0\\ 0& other\end{array}---\left(22\right)$

Definition status estimation difference isOutput error is

Assumed condition 1 meets, then there is linear transformation battle array T so that (A_a,M,C_a) become the form of formula (23):

${\mathrm{TA}}_a{T}^{-1}=\left[\begin{array}{cc}{A}_{a1}& A_{a2}\\ A_{a3}& A_{a4}\end{array}\right],TM=\left[\begin{array}{c}0\\ M_2\end{array}\right],C_a{\left(T\right)}^{-1}=\[\begin{array}{cc}0& T\end{array}\]---\left(23\right)$

In formula, T is an orthogonal matrix,M₀It is a non-singular matrix, (A_a22,A_a31) it is the most observable.

By formula (18)-(19) and (20)-(21) can systematic error dynamical equation be:

$\stackrel{\·}{e}\left(t\right)=A_{a0}e\left(t\right)+G_n{v}_{\mathrm{\ρ}}-{\mathrm{Mf}}_{as}---\left(24\right)$

In formula, A_a0=A_a-G_lC_a.Observer matrix can be taken asL=[L₁ 0]。

Output dynamic error is:

${\stackrel{\·}{e}}_{ya}\left(t\right)=C_a{A}_{a0}e\left(t\right)+C_a{G}_n{v}_{\mathrm{\ρ}}-C_a{\mathrm{Mf}}_{as}---\left(25\right)$

When sliding formwork motion arrivesDiscontinuous term can be expressed as follows

v_ρ=-(C_aG_n)^-1(C_aA_a0e(t)-C_aMf_as(t)) (26)

Then

E (t)=(I-G_n(C_aG_n)^-1C_a)A_aoe(t)-C_aMf_as(t) (27)

In formula,

Reconstruct to utilize designed sliding mode observer to realize actuator and sensor failures simultaneously, define linear coordinate Conversion: T_L:e→e_L, takeThen matrix (A_a,M,C_a) translate into following form:

$A_{aL}=\left[\begin{array}{cc}{A}_{a1L}& A_{a2L}\\ A_{a3L}& A_{a4L}\end{array}\right],M_L=\left[\begin{array}{c}0\\ M_{2L}\end{array}\right],C_{aL}=\[\begin{array}{cc}0& I_p\end{array}\]---\left(28\right)$

In formula, A_a1L=A_a1+LA_a3, M_2L=TM₂Being all non-singular matrix, non-linear gain matrix meets Liapunov side Journey, desirable G_nL=[0 I_p]^Τ

Under this coordinate transform, state and the output bias equation of system be:

${\stackrel{\·}{e}}_{a1}\left(t\right)=A_{a1L}{e}_1\left(t\right)+(A_{a1L}-G_{l1L})e_{ya}\left(t\right)---\left(29\right)$

${\stackrel{\·}{e}}_{ay}\left(t\right)=A_{a3L}{e}_{a1}\left(t\right)+(A_{a3L}-G_{12L})e_{ay}\left(t\right)-M_{2L}{f}_{as}\left(t\right)+v_{\mathrm{\ρ}}---\left(30\right)$

Assume that sliding formwork motion arrives,A_a1LStable, then e_a1(t)→0。

Therefore, can obtain

${\stackrel{\·}{e}}_{ay}\left(t\right)=-M_{2L}{f}_{as}\left(t\right)+v_{\mathrm{\ρ}}=0---\left(31\right)$

Then the reconstruction value of unknown failure vector is:

${\hat{f}}_{as}\left(t\right)={\left(M_{2L}^T{M}_{2L}\right)}^{-1}{M}_{2L}^T{v}_{\mathrm{\ρ}}---\left(32\right)$

Continuous function is in like manner utilized to represent discontinuous term:

$v_{\mathrm{\ρ}eq}=-\mathrm{\ρ}\left|\right|M_{2L}\left|\right|\frac{P_2{e}_{ay}}{\left|\right|P_2{e}_{ay}+\mathrm{\δ}\left|\right|}---\left(33\right)$

In formula, σ is that scalar that value is less is to ensure the precision of failure reconfiguration.

Here the stability of observer and state in finite time, arrive sliding-mode surface can be by setting up Liapunov letter Number proves, observer gain matrix can be sent out by POLE PLACEMENT USING and solve.Same available given failure reconfiguration Error threshold, judges failure reconfiguration precision.5th step: the faults-tolerant control of wind generator system

Theoretical by above failure reconfiguration, it is possible to obtain the information such as fault waveform, amplitude in real time, it is to avoid to produce and comment The complexity of valency residual signals.Hence with sensor fault reconstruction approach base valueTo sensor fault output signal y_fIt is corrected, then System output calibration signal is:

y_r=y_f-Ff_s (34)

Then output calibration signal y is utilized_rInput as controller, it is ensured that system output is controlled by controller accurately Amount u, and then realize sensor fault apparatus for lower wind electricity generation system active tolerant control.Active tolerant control strategy designed by paper The structure and parameter of original system output feedback controller, the controller still using system originally to design need not be changed, utilize defeated Go out correction signal y_rSystem is carried out faults-tolerant control, fault type is made without anticipation, highlight active tolerant control energy Enough process the ability of unknown failure.Now controller input is

u_c=y_f-Ff_s (35)

In formula, u_c=[Ω_h Γ_G]^ΤInput for controller.

In like manner actuator failures be may be used without the faults-tolerant control being similar to.

Accompanying drawing explanation

The basic structure of Fig. 1 wind generator system based on double fed induction generators (DFIG)

Fig. 2 membership function structure chart

Fig. 3 wind generator system faults-tolerant control structure chart

Fig. 4 wind generator system HWIL simulation control principle drawing

Specific implementation method

The invention will be further described with example below in conjunction with the accompanying drawings.

The first step: set up the Global fuzzy model of wind generator system

According to the state equation of the wind generator system obtained by formula (5), T-S fuzzy algorithmic approach is utilized to obtain wind-power electricity generation system Shown in the overall T-S fuzzy model such as formula (8) of system.

Second step: wind generator system multi-fault Diagnosis and reconstruction strategy

By to actuator failures and sensor fault characteristic analysis, actuator failures is internal system fault, and it can be led Cause system relevant parameter changes, and eventually affects the output of system, and therefore it can be considered sensor fault, passes through sensor It is theoretical that hardware redundancy technology combines sliding mode observer, sets up logic residual error table, it is achieved multi-fault Diagnosis.

Secondly by introduce a simple wave filter, sensor fault is converted into actuator failures, set up one by Original actuator failures and the virtual actuator failures of sensor fault composition, by coming the reconstruct of virtual actuator failures Realize two kinds of faults to reconstruct simultaneously.

3rd step: the faults-tolerant control of wind generator system

Theoretical by above failure reconfiguration, it is possible to obtain the information such as fault waveform, amplitude in real time, it is to avoid to produce and comment The complexity of valency residual signals.Hence with sensor fault reconstruction approach base valueTo sensor fault output signal y_fIt is corrected, then System output calibration signal is:

y_r=y_f-Ff_s (34)

Then output calibration signal y is utilized_rInput as controller, it is ensured that system output is controlled by controller accurately Amount u, and then realize sensor fault apparatus for lower wind electricity generation system active tolerant control.Active tolerant control strategy designed by paper The structure and parameter of original system output feedback controller, the controller still using system originally to design need not be changed, utilize defeated Go out correction signal y_rSystem is carried out faults-tolerant control, fault type is made without anticipation, highlight active tolerant control energy Enough process the ability of unknown failure.Now controller input is

u_c=y_f-Ff_s (35)

In formula, u_c=[Ω_h Γ_G]^ΤInput for controller.

In like manner actuator failures be may be used without the faults-tolerant control being similar to.

The wind generator system of multiple faults described above is diagnosed with fault-tolerant control module in HWIL simulation by the 4th step Realize on device AD5436.Its input is digital quantity signal corresponding to the wind velocity signal of wind energy conversion system, is output as tach signal and electricity Magnetic torque signal is for carrying out closed loop feedback control to wind generator system.PC is connected with AD5436 by Ethernet, passes through Analog quantity is changed into digital quantity by AD/DA changer.When wind speed variable reaches gear-box, driving-chain drive motor rotates, logical Cross AC-AC conversion, deliver to electrical network.

The present invention is in the Shunt-connected Wind Power Generation System, as shown in the figure.Main modular has wind energy conversion system, gear-box, double-fed Influence generator, converters, electric capacity, transformator and electrical network etc..The fault will built in software Matlab Diagnosis copies in HWIL simulation workbox with the model of faults-tolerant control, designs in conjunction with associate power electronic section, carries out half real Thing emulates, and utilizes the waveform of the performance parameter power coefficient after the faults-tolerant control of HWIL simulation case gained and tip speed ratio Figure, contrasts with the situation under fault, the effectiveness of checking fault-tolerant controller designed by the present invention.

During fan operation, failure diagnosis unit involved in the present invention can obtain accurate fault letter by real-time online Breath, compensates the fault of wind generator system, does not change the structure and parameter of whole wind power system and controller, therefore Controller design is simple and convenient.After fault occurs, the active tolerant control device designed by utilizing, whole system remains able to Maintain and original performance indications similar level, and stable operation, complete corresponding control task.

Claims (1)

1. wind generator system multi-fault Diagnosis and faults-tolerant control, is characterized in that:

Wind generator system multi-fault Diagnosis strategy is proposed；

Sensor fault, as a example by high speed shaft speed probe, uses hardware redundancy technology to export high speed shaft rotating speed Ω in system_hLetter Number position install two duplicate speed probes be used for measuring high speed shaft rotating speed, be designated as Ω_h,m1And Ω_h,m2, i=1, 2；When system is without sensor fault, Ω_h,_m1=Ω_h,m2=Ω_h；When system generation sensor fault, signal is Ω_h,m1= Ω_h+ΔΩ_h,m1, Ω_h,m2=Ω_h+ΔΩ_h,m2；Additional Δ Ω_h,m1With Δ Ω_h,m2Represent speed probe 1 and revolution speed sensing The contingent fault value of device 2；

By comparing the measured value between two same type of sensors of same position, it is achieved wind generator system fault diagnosis Purpose；Difference between these two signals can directly show the generation of fault, if | | Ω_h,m1-Ω_h,m2| | in certain time period Inside it is not equal to 0, then system jam is described；

T-S fuzzy-sliding-mode observer can accurately draw high speed shaft speed estimate valueIt is before any sensor fault occurs After the most closely actual value；Actuator failures is internal system fault, and it can cause system relevant parameter to change, and affects system Output, can be considered sensor fault；NoteWithMake residual signals, system fault diagnosis strategy Specifically describe as shown in table 1:

Table 1 multi-fault Diagnosis logical table

Actuator and sensor failures is reconstructed simultaneously；

Consider sensor fault and the simultaneous system of actuator failures (12), it is assumed that system (12) meetsAnd the constant zero point of system is stable；Use orthogonal matrix T_RThe system output square being multiplied by formula (12) Battle array, can obtain formula (13) and formula (14):

$y_1\left(t\right)={\stackrel{\‾}{C}}_{1}x\left(t\right)---\left(13\right)$

$y_2\left(t\right)={\stackrel{\‾}{C}}_{2}x\left(t\right)+F_1{f}_s\left(t\right)---\left(14\right)$

Introduce wave filter, define one group of new state variable z (t) as the output y containing sensor fault₂(t) filtered letter Number, such as formula (15):

$\stackrel{\·}{z}\left(t\right)=-A_{f}z\left(t\right)+A_f{y}_2\left(t\right)=-A_{f}z\left(t\right)+A_f{\stackrel{\‾}{C}}_{2}x\left(t\right)+A_f{F}_1{f}_s\left(t\right)---\left(15\right)$

In formula, A_fIt it is a stationary filtering matrix；

Simultaneous formula (12) can obtain to (15):

$\left[\begin{array}{c}\stackrel{\·}{x}\left(t\right)\\ \stackrel{\·}{z}\left(t\right)\end{array}\right]=\left[\begin{array}{cc}\stackrel{\‾}{A}& 0\\ A_f{\stackrel{\‾}{C}}_2& -A_f\end{array}\right]\left[\begin{array}{c}x\left(t\right)\\ z\left(t\right)\end{array}\right]+\left[\begin{array}{c}\stackrel{\‾}{B}\\ 0\end{array}\right]u\left(t\right)+\left[\begin{array}{cc}D& 0\\ 0& A_f{F}_1\end{array}\right]\left[\begin{array}{c}{f}_a\left(t\right)\\ f_s\left(t\right)\end{array}\right]---\left(16\right)$

$\left[\begin{array}{c}{y}_1\left(t\right)\\ z\left(t\right)\end{array}\right]=\left[\begin{array}{cc}{\stackrel{\‾}{C}}_1& 0\\ 0& I_h\end{array}\right]\left[\begin{array}{c}x\left(t\right)\\ z\left(t\right)\end{array}\right]---\left(17\right)$

Consideration formula (16), (17), then wind generator system can be expressed as again:

${\stackrel{\·}{x}}_a\left(t\right)=A_a{x}_a+B_{a}u\left(t\right)+{\mathrm{Mf}}_{as}\left(t\right)---\left(18\right)$

y_a(t)=C_ax_a(t) (19)

In formula,Represent virtual Actuator failures vector, it is norm-bounded, i.e. f_as(t)≤η(x,u,t)+η₀, η₀It is a normal number, x_a(t)=[x (t )^Τ z(t)^Τ]^ΤFor the system state variables after coordinate transform；

After the above-mentioned conversion to system (12), new system (18), (19) do not contain sensor fault, but will sensing Device fault all becomes actuator failures and processes with actuator failures.Design corresponding sliding formwork Failure Observer and combine of equal value defeated Go out to inject theory, virtual faults reconstruction value can be obtainedSolve two kinds of faults f_s(t)、f_aReconstruction while (t)；

The design process of brief description wind generator system active tolerant control device:

The existence condition of T-S fuzzy-sliding-mode observer:

(1)rank(C_aM)=rank (M)；

(2)(A_a,M,C_a) constant zero point be stable；

The form of design wind generator system sliding mode observer is as follows:

${\stackrel{\·}{\hat{x}}}_a\left(t\right)=A_a{\hat{x}}_a+B_{a}u-G_l{e}_y+G_n{v}_{\mathrm{\ρ}}---\left(20\right)$

${\hat{y}}_a=C_a{\hat{x}}_a---\left(21\right)$

In formula, ρ is correction factor (gain), fault amplitude determined；G_l, G_nFor Robust Sliding Mode Observer treat constant linear instead Feedback matrix and nonlinear feedback matrix.In order to guarantee sliding formwork motion and design sliding mode control strategy v_ρ, i.e.

$v_{\mathrm{\ρ}}=\left\{\begin{array}{cc}-\mathrm{\ρ}\left|\right|M_{2L}\left|\right|\frac{P_2{e}_{ya}}{\left|\right|P_2{e}_{ya}\left|\right|}& e_{ya}\≠0\\ 0& other\end{array}\right.---\left(22\right)$

Definition status estimation difference isOutput error is

Assumed condition 1 meets, and there is linear transformation battle array T so that (A_a,M,C_a) become the form of formula (23):

${\mathrm{TA}}_a{T}^{-1}=\left[\begin{array}{cc}{A}_{a1}& A_{a2}\\ A_{a3}& A_{a4}\end{array}\right],TM=\left[\begin{array}{c}0\\ M_2\end{array}\right],C_a{\left(T\right)}^{-1}=\left[\begin{array}{cc}0& T\end{array}\right]---\left(23\right)$

In formula, T is an orthogonal matrix,M₀It is a non-singular matrix, (A_a22,A_a31) it is the most observable；

By formula (18)-(19) and (20)-(21) can systematic error dynamical equation be:

$\stackrel{\·}{e}\left(t\right)=A_{a0}e\left(t\right)+G_n{v}_{\mathrm{\ρ}}-{\mathrm{Mf}}_{as}---\left(24\right)$

In formula, A_a0=A_a-G_lC_a；Observer matrix can be taken as

Output dynamic error isWork as sliding formwork When motion arrivesDiscontinuous term can be expressed as follows

v_ρ=-(C_aG_n)^-1(C_aA_a0e(t)-C_aMf_as(t)) (26)

Then

E (t)=(I-G_n(C_aG_n)^-1C_a)A_aoe(t)-C_aMf_as(t) (27)

In formula,

Reconstructing to utilize designed sliding mode observer to realize actuator and sensor failures, definition linear coordinate becomes simultaneously Change: T_L:e→e_L, takeThen matrix (A_a,M,C_a) translate into following form:

$A_{aL}=\left[\begin{array}{cc}{A}_{a1L}& A_{a2L}\\ A_{a3L}& A_{a4L}\end{array}\right],M_L=\left[\begin{array}{c}0\\ M_{2L}\end{array}\right],C_{aL}=\left[\begin{array}{cc}0& I_p\end{array}\right]---\left(28\right)$

In formula, A_a1L=A_a1+LA_a3, M_2L=TM₂Being all non-singular matrix, non-linear gain matrix meets Lyapunov Equation, takes

Under this coordinate transform, state and the output bias equation of system be:

${\stackrel{\·}{e}}_{a1}\left(t\right)=A_{a1L}{e}_1\left(t\right)+(A_{a1L}-G_{11L})e_{ya}\left(t\right)---\left(29\right)$

${\stackrel{\·}{e}}_{ay}\left(t\right)=A_{a3L}{e}_{a1}\left(t\right)+(A_{a3L}-G_{12L})e_{ay}\left(t\right)-M_{2L}{f}_{as}\left(t\right)+v_{\mathrm{\ρ}}---\left(30\right)$

Assume that sliding formwork motion arrives,A_a1LStable, then e_a1(t)→0；

Therefore, can obtain

${\stackrel{\·}{e}}_{ay}\left(t\right)=-M_{2L}{f}_{as}\left(t\right)+v_{\mathrm{\ρ}}=0---\left(31\right)$

Then the reconstruction value of unknown failure vector is:

${\hat{f}}_{as}\left(t\right)={\left(M_{2L}^T{M}_{2L}\right)}^{-1}{M}_{2L}^T{v}_{\mathrm{\ρ}}---\left(32\right)$

Continuous function is in like manner utilized to represent discontinuous term:

$v_{\mathrm{\ρ}eq}=-\mathrm{\ρ}\left|\right|M_{2L}\left|\right|\frac{P_2{e}_{ay}}{\left|\right|P_2{e}_{ay}+\mathrm{\δ}\left|\right|}---\left(33\right)$

In formula, σ is that scalar that value is less is to ensure the precision of failure reconfiguration.