CN104808662A - Control method for suppressing ship course disturbance based on data driving - Google Patents

Abstract

The invention belongs to the fields of ship engineering, control science and control engineering, and particularly relates to a control method for suppressing ship course disturbance based on data driving. The method is realized in a way that the input end of a residual error generator module based on the full-state observer class receives actual ship navigation yawing and rolling angle signals and control signals and outputs rolling angle residual error signals and yawing angle residual error signals; the two types of residual error signals and course expected values are processed by a Youla parameterization controller; and the output end of the Youla parameterization controller is connected with the input end of a steering engine servo system. The control method of data driving is fully utilized, and wave disturbance is effectively suppressed under the premise of maintaining no off-course of ship motion. The control method is simple in structure and easy to realize so that the requirements of actual engineering applications can be greatly met.

Description

A kind of control method of the suppression ship course disturbance based on data-driven

Technical field

The invention belongs to Marine engineering, control science and control engineering field, be specifically related to a kind of control method of the suppression ship course disturbance based on data-driven.

Background technology

When naval vessel rides the sea, the impact of various sea situation complicated and changeable can be subject to.The antijamming capability how improving Ship System is Chinese scholars question of common concern.The control method of research both at home and abroad comprises traditional PI D, neural network control method, LQG control algolithm, FUZZY ALGORITHMS FOR CONTROL, optimal control algorithm, PREDICTIVE CONTROL etc. at present.

In practical engineering application, conventional linear ship model CONTROLLER DESIGN, designs linear controller after therefore needing to carry out linearization to nonlinear model, then applies it in the middle of nonlinear model.Utilize the direct identification equivalent space of off-line data and vector, and then build residual signals, build the optimal controller based on residual error.And on the basis of the equivalent space of identification, further by the Youla Parameterization Controller of data-driven method design based on observer.When carrying out controller optimization, employing be residual error Tuning, the method is the Grad utilizing residual error data direct estimation quadratic performance index, uses Gradient Descent vehicle air-conditioning device.In ship's navigation process, use the controller after on-line optimization effectively to kill the sea the impact of interference and mission nonlinear, improve the robustness of system.

Summary of the invention

The object of the present invention is to provide a kind of sea wind, wave of can effectively suppressing to the disturbing influence of ship's navigation, response rapidly, the control method of the suppression ship course disturbance based on data-driven of good, the strong robustness of control performance.

The present invention is achieved in that

(1) the Residual Generation device module input based on state observer class receives actual ship's navigation bow, roll angle signal, control signal u, exports roll angle residual signals with yaw angle residual signals r _ψ:

State observer is $\hat{x}(k+1)=A_x\hat{x}\left(k\right)+B_{x}u\left(k\right)+L_{x}y\left(k\right),$ Residual Generation device is wherein for the state estimation matrix of state observer, the residual matrix that r (k) is state observer, A _x, B _x, L _x, G, C _x, D _xfor the matrix of coefficients of state observer, u (k) is ordered rudder angle, the yaw angle that y (k) is boats and ships real navigation and roll angle;

(2) two kinds of residual signals and course expectation value are through the process of Youla Parameterization Controller:

Youla Parameterization Controller is wherein u (z) is gating matrix, y _dz course value that () sets for boats and ships, r (z) is residual matrix, and Q (z) is Youla parameterization matrix, and F is state gain matrix;

(3) Youla Parameterization Controller output terminal connects the input end of Rudder Servo System, the control signal that the output terminal of Rudder Servo System exports is input in ship horizontal motion reference model module, the actual ship bow that ship horizontal motion reference model module exports, rolling signal and controller output signal are linked in the Residual Generation device module based on state observer class simultaneously.

Beneficial effect of the present invention is:

The present invention makes full use of the control method of data-driven, under the prerequisite keeping ship motion not go off course, effectively inhibits sea wave disturbance.

Patent structure of the present invention is simple, is easy to realize, and can meet the needs of practical engineering application very well.It is 20 ° at given desired course angle, when the disturbance of three grades of waves, the rolling residual sum yawing residual error exported by Youla parameter controller is compared with the output residual values not adding Youla parameter controller, effect has had obvious improvement, as shown in Figure 1, the validity of this algorithm is demonstrated.The present invention effectively can suppress the disturbance of three grades of waves of simulating, and rolling motion can be effectively suppressed, as shown in Figure 2.

Accompanying drawing explanation

The residual error simulated effect figure that Fig. 1 Youla Parameterization Controller controls;

The Output simulation design sketch that Fig. 2 Youla Parameterization Controller controls;

The residual error simulated effect figure that Fig. 3 Youla Parameterization Controller controls;

The Output simulation design sketch that Fig. 4 Youla Parameterization Controller controls;

Fig. 5 control system composition frame chart of the present invention;

Fig. 6 control method process flow diagram of the present invention.

Embodiment

Below in conjunction with accompanying drawing, the present invention is described further.

Based on the control method of the suppression ship course disturbance of data-driven, this system comprises Residual Generation device module, Youla Parameterization Controller module, Rudder Servo System module, ship horizontal motion module based on state observer class.State observer is $\hat{x}(k+1)=A_x\hat{x}\left(k\right)+B_{x}u\left(k\right)+L_{x}y\left(k\right),$ Residual Generation device is $r\left(k\right)=\mathrm{Gy}\left(k\right)-C_x\hat{x}\left(k\right)-D_{x}u\left(k\right),$ Wherein for the state estimation matrix of state observer, the residual matrix that r (k) is state observer, A _x, B _x, L _x, G, C _x, D _xfor the matrix of coefficients of state observer, u (k) is ordered rudder angle, the yaw angle that y (k) is boats and ships real navigation and roll angle.Youla Parameterization Controller is wherein u (z) is gating matrix, y _dz course value that () sets for boats and ships, r (z) is residual matrix, and Q (z) is Youla parameterization matrix, F is state gain matrix, it needs to be configured according to controllable pair (A, B), and fundamental purpose is exactly to ensure system stability.Residual Generation device module based on observer class can observe yaw angle ψ, roll angle residual signals r, its input end receives actual ship's navigation bow, roll angle signal, control signal u, exports roll angle residual signals with yaw angle residual signals r _ψtwo kinds of residual signals and course expectation value are again after the process of Youla Parameterization Controller, output terminal connects the input end of Rudder Servo System, the control signal that its output terminal exports is input in ship horizontal motion reference model module, the actual ship bow exported, rolling signal and controller output signal are linked in the Residual Generation device based on observer class simultaneously, thus constitute a close loop control circuit, utilize programming software Matlab to realize computing machine and control.According to the composition of actual ship horizontal motion control system, Residual Generation device is embedded in state observer, online acquisition input, output and residual signals, calculate input, output signal Grad to youla parameter, then the Grad of quadratic performance index J and J thereof is calculated, calculate the youla parameter value of i+1 time according to gradient descent method, until the convergence of quadratic performance index J value, the youla parameter value obtained is only optimal value.Now designed Youla controller effectively could kill the sea model to the interference produced in ship running process.

Fig. 2 is 15 ° at given desired course angle, when level Four wave disturbance, the rolling residual sum yawing residual error exported by Youla parameter controller is compared with the output residual values not adding Youla parameter controller, as shown in Figure 3, effect also improves significantly, and corresponding roll angle exports and yaw angle exports as shown in Figure 4.Fig. 4 by two groups of sea situation emulation experiments, demonstrates the validity of this algorithm, embody the present invention have be easy to modeling, response rapidly, control performance better, strong robustness and the advantage such as logical organization is simple.

Composition frame chart of the present invention as shown in Figure 5, is made up of the simulation analysis module of the analog module of Rudder Servo System, ship horizontal motion reference model module, sea wave disturbance module, Residual Generation device module, Youla Parameterization Controller module and control effects based on state observer.A kind of workflow diagram of control method of the suppression ship course disturbance based on data-driven as shown in Figure 6.Above-mentioned module utilizes programming software Matlab to be realized by computer programming.According to the composition of actual ship horizontal motion control system, Residual Generation device is embedded in state observer, online acquisition input, output and residual signals, calculate input, output signal Grad to youla parameter, then the Grad of quadratic performance index J and J thereof is calculated, calculate the youla parameter value of i+1 time according to gradient descent method, until the convergence of quadratic performance index J value, the youla parameter value obtained is only optimal value.Now designed Youla controller effectively could kill the sea model to the interference produced in ship running process.

Realize technical scheme of the present invention:

Based on a control method for the suppression ship course disturbance of data-driven, it is characterized in that comprising the following steps:

Parity space method is utilized to design Residual Generation device

Parity space method is one comparatively simple and effective Residual Generation device method for designing, it checks the equivalence (i.e. consistance) of diagnosis object mathematical relation by the actual value of system input, output (or part exports), thus reach detection and the object be separated.Consider the Minimal Realization of the state equation model of the considerable controlled discrete linear systems of a class

x(k+1)＝Ax(k)+Bu(k)+w(k)

y(k)＝Cx(k)+Du(k)+v(k) (1)

Wherein, u ∈ R ^lrepresent control inputs, y ∈ R ^mrepresent to measure and export, x ∈ R ⁿrepresent system state.W ∈ R ⁿrepresent systematic procedure noise, v ∈ R ^mrepresent systematic survey noise, assuming that systematic procedure noise and systematic survey noise are mutual uncorrelated white noise, respective variance is designated as R _w, R _v.A, B, C, D are the system matrixes of corresponding dimension.Assuming that system matrix A, B, C, D, systematic education n and noise information R _w, R _vequal the unknown, system is under unfaulty conditions, carry out design Residual Generation device.

When systematic parameter and noise information the unknown, by open loop acquisition system input/output variable, set up two groups of typical data ordering structures, namely Hankel matrix can be expressed as:

Wherein, U _frepresent following input Hankel matrix, the U _prepresent input Hankel matrix, Y in the past _frepresent following output Hankel matrix, the Y _prepresent output Hankel matrix in the past, W _p, W _frepresent the Hankel matrix of noise, N represents sample size, and s is systematic education, s _f=s+1, meets s _f> n.So, the matrix model of Minimal Realization system can be described as:

Y _f＝Γ _sX(i)+H _s,uU _f+H _s,wW _f+V _f(2)

Wherein, system state matrix X (i)=[x (i) x (i+1) ... x (i+N-1)] ∈ R ^{n × N}, expansion observing matrix ${\mathrm{\Γ}}_s\∈R^{{\mathrm{ms}}_f\×n}$ With relevant Markov matrix $H_{s,u}\∈R^{{\mathrm{ms}}_f\×{\mathrm{ls}}_f}$ With $H_{s,w}\∈R^{{\mathrm{ms}}_f\×{\mathrm{ns}}_f}$ Be expressed as

Γ _s＝[C CA … CA ^s] ^T

Input and output matrix equation (2) belongs to the Subspace Identification pattern of standard.The object of its Subspace Identification is the interference overcoming noise.So, conventional data structure Z _fand Z _pbe expressed as

$Z_f=\left[\begin{array}{c}{Y}_f\\ U_f\end{array}\right]\∈R^{(l+m)s_f\×N},Z_p=\left[\begin{array}{c}{Y}_p\\ U_p\end{array}\right]\∈R^{(l+m)s_f\×N}$

If there is vector meet v _sΓ _s=0, wherein, Γ _s=[C CA ... CA ^s] ^t; So, v is defined _sfor the equal vector under exponent number s.Residual signals r (k) of parity space method is defined as

r(k)＝v _sy _f-v _sH _s,uu _f(3)

For realizing the data-driven design of (3) formula Residual Generation device, first based on Z _fand Z _pbuild

$\frac{1}{N}{Z}_f{Z}_p^T=\frac{1}{N}\left[\begin{array}{c}{Y}_f\\ Y_p\end{array}\right]Z_p^T=\frac{1}{N}\left[\begin{array}{cc}{\mathrm{\Γ}}_s& H_{u,s}\\ 0& I\end{array}\right]\left[\begin{array}{c}X\left(i\right)\\ U_f\end{array}\right]Z_p^T+\frac{1}{N}\left[\begin{array}{c}{H}_{s,w}{W}_f\\ 0\end{array}\right]Z_p^T$

In formula, the data Z in " past " _pwith the noise H in future _w,sw _f+ V _funcorrelated mutually, therefore can obtain further

$\frac{1}{N}\left[\begin{array}{c}{H}_{s,w}{W}_f\\ 0\end{array}\right]Z_p^T\≈0\⇒\frac{1}{N}{Z}_f{Z}_p^T\≈\frac{1}{N}\left[\begin{array}{cc}{\mathrm{\Γ}}_s& H_{u,s}\\ 0& I\end{array}\right]\left[\begin{array}{c}X\left(i\right)\\ U_f\end{array}\right]Z_p^T$

Definition Γ _scomplementary space (or orthogonal set) meet be systematic education s equivalent space.

Complete algorithm step based on parity space method design Residual Generation device is as follows:

Image data Z _fand Z _p, build

Right carry out SVD decomposition:

$\frac{1}{N}{Z}_f{Z}_p^T=U_z\left[\begin{array}{cc}{\mathrm{\Σ}}_{z,1}& \\ & 0\end{array}\right]V_z^T,U_z=\left[\begin{array}{cc}{U}_{z,11}& U_{z,12}\\ U_{z,21}& U_{z,22}\end{array}\right]$

Wherein, $U_{z,12}\∈R^{{\mathrm{ms}}_f\×({\mathrm{ms}}_f-n)},U_{z,22}\∈R^{{\mathrm{ls}}_f\×({\mathrm{ls}}_f-n)}.$

Arrange ${\mathrm{\Γ}}_s^{\⊥}=U_{z,12}^T,{\mathrm{\Γ}}_s^{\⊥}{H}_{s,u}=-U_{z,22}^T$

Build the Residual Generation device (3 formula) based on parity space method

The design of observer class Residual Generation device

Complicated by method comparison during (3) formula known parity space method structure residual error, and relatively just fairly simple based on the Residual Generation device of observer class.Therefore the observer class Residual Generation device constructed further based on data-driven is needed.As the model matrix A of system, B, C, D are known, can build observer Residual Generation device, be embodied in by Luenberger system of equations:

TA-A _zT＝LC，c _zT＝gC，B _z＝TB-LD，d _z＝gD

The observer class Residual Generation device of structure is:

z(k+1)＝A _zz(k)+B _zu(k)+Ly(k)∈R ^s

(4)

r(k)＝gy(k)-c _zz(k)-d _zu(k)∈R

Wherein, the corresponding matrix of coefficients of the observer of structure is A _z∈ R ^{s × s}(eigenwert is in the unit circle taking initial point as the center of circle, stable in system), B _z∈ R ^{s × l}, c _z∈ R ^{l × s}, g _z∈ R ^{l × m}, L ∈ R ^{s × m}, T ∈ R ^{s × n}.This maker can generate residual error online, according to residual error design optimization controller, realizes the control object suppressing ship course disturbance.

According to the relation of equivalence of equivalent space vector with observer class Residual Generation device, if namely given equivalent space is vectorial

v _s＝[v _s,0v _s,1… v _s,s]∈R ^l×m,i＝0,1,…,s

So, matrix parameter and equal vector v _sand v _sh _u,sthere is following relation:

$L=\left[\begin{array}{c}{v}_{s,0}\\ v_{s,1}\\ .\\ .\\ .\\ v_{s,s-1}\end{array}\right]$

c _z＝[0 0 … 1]，g＝v _s,s，

$B_z=\left[\begin{array}{c}{B}_{z,1}\\ B_{z,2}\\ .\\ .\\ .\\ B_{z,2}\end{array}\right]=\left[\begin{array}{c}{t}_{1}B-v_{s,0}D\\ t_{2}B-v_{s,1}D\\ .\\ .\\ .\\ t_{s}B-v_{s,s-1}D\end{array}\right]=\left[\begin{array}{c}{v}_s{H}_{s,0}\\ v_s{H}_{s,1}\\ .\\ .\\ .\\ v_s{H}_{s,s-1}\end{array}\right]$

Investigate m and tie up observer class Residual Generation device group, wherein z (k) state matrix that is observer, its exponent number ms significantly may be greater than n, is therefore difficult to online observation and state estimation, therefore needs the observer on design n rank,

$z^i(k+1)=A_z{z}^i\left(k\right)+B_z^{i}u\left(k\right)+L^{i}y\left(k\right)$

(5)

$r^i\left(k\right)=g^{i}y\left(k\right)-c_z{z}^i\left(k\right)-d_z^{i}u\left(k\right)$

First the mDos in (5) is written as compact form

Wherein

$G=\left[\begin{array}{c}{v}_{s,s}^1\\ .\\ .\\ .\\ v_{s,s}^m\end{array}\right]=\left[\begin{array}{c}{g}_1\\ .\\ .\\ .\\ g_m\end{array}\right],$

Carry out matrixing to have

New state matrix A can be released _xcomprise A _zn eigenwert, these eigenwerts are also all zero.So, other observer matrix is converted further and can be obtained

Thus, n rank state observer and m dimension Residual Generation device can be constructed

$\hat{x}(k+1)=A_x\hat{x}\left(k\right)+B_{x}u\left(k\right)+L_{x}y\left(k\right)$

(9)

$r\left(k\right)=\mathrm{Gy}-C_x\hat{x}\left(k\right)-D_{x}u\left(k\right)$

Wherein for the state estimation matrix of state observer, the residual matrix that r (k) is state observer, A _x, B _x, L _x, G, C _x, D _xfor the matrix of coefficients of state observer, u (k) is ordered rudder angle, the yaw angle that y (k) is boats and ships real navigation and roll angle.

Design Youla Parameterization Controller

By $r\left(k\right)=\mathrm{Gy}\left(k\right)-C_x\hat{x}\left(k\right)-D_{x}u\left(k\right)$ The output that can obtain system is

$y=G^{-1}{C}_x\hat{x}+G^{-1}{D}_{x}u\left(k\right)---\left(10\right)$

So state observer (9) formula also can be described as following open cycle system

$\hat{x}(k+1)=(A_x+L_x{G}^{-1}{C}_x)\hat{x}\left(k\right)+(B_x+L_x{G}^{-1}{D}_x)u\left(k\right)$

(11)

$y\left(k\right)=G^{-1}{C}_x\hat{x}\left(k\right)+G^{-1}{D}_{x}u\left(k\right)$

System matrix A, B, C, D can be expressed as

A＝A _x+L _xG ^-1C _x，B＝B _x+L _xG ^-1D _x，C＝G ^-1C _x，D＝G ^-1D _x(12)

So ssystem transfer function G _uz the left and right coprime factorization of () and stability controller K (z) is expressed as

$G_u\left(z\right)=N\left(z\right)M^{-1}\left(z\right)={\hat{M}}^{-1}\left(z\right)\hat{N}\left(z\right),N\left(z\right),M\left(z\right),\hat{N}\left(z\right),\hat{M}\left(z\right)\∈{\mathrm{RH}}_{\∞}---\left(13\right)$

$K\left(z\right)=Y\left(z\right)X^{-1}\left(z\right)={\hat{X}}^{-1}\left(z\right)\hat{Y}\left(z\right),Y\left(z\right),X\left(z\right),\hat{Y}\left(z\right),\hat{X}\left(z\right)\∈{\mathrm{RH}}_{\∞}---\left(14\right)$

Wherein,

M(z)＝I+F(zI-A _F)B,A _F＝A+BF

N(z)＝D+C _F(zI-A _F) ^-1B,C _F＝C+DF

Y(z)＝F(zI-A _F) ^-1L

X(z)＝I+C _F(zI-A _F) ^-1L

$\hat{M}\left(z\right)=I-C{(\mathrm{zI}-A_L)}^{-1}L,A_L=A-\mathrm{LC}$

$\hat{N}\left(z\right)=D+C(\mathrm{zI}-A_L)B_L,B_L=B-\mathrm{LD}$

$\hat{Y}\left(z\right)=F{(\mathrm{zI}-A_L)}^{-1}L$

$\hat{X}\left(z\right)=I-F{(\mathrm{zI}-A_L)}^{-1}{B}_L$

Wherein gain F and L ensures that A+BF and A-LC stablizes.Matrix M (z), N (z), X (z), Y (z) and meet Bezout identical relation

$\left[\begin{array}{cc}\hat{X}\left(z\right)& -\hat{Y}\left(z\right)\\ -\hat{N}\left(z\right)& \hat{M}\left(z\right)\end{array}\right]\left[\begin{array}{cc}M\left(z\right)& Y\left(z\right)\\ N\left(z\right)& X\left(z\right)\end{array}\right]=\left[\begin{array}{cc}M\left(z\right)& Y\left(z\right)\\ N\left(z\right)& X\left(z\right)\end{array}\right]\left[\begin{array}{cc}\hat{X}\left(z\right)& -\hat{Y}\left(z\right)\\ -\hat{N}\left(z\right)& \hat{M}\left(z\right)\end{array}\right]=\left[\begin{array}{cc}I& 0\\ 0& I\end{array}\right]---\left(15\right)$

Based on G _uz () left and right coprime factorization, Youla parameterization can pass through Youla matrix Q (z) design stability control law

$\begin{array}{c}K\left(Q\left(z\right)\right)=(Y\left(z\right)-M\left(z\right)Q\left(z\right)){(X\left(z\right)-N\left(z\right)Q\left(z\right))}^{-1}\\ ={(\hat{X}\left(z\right)-Q\left(z\right)\hat{N}\left(z\right))}^{-1}(\hat{Y}\left(z\right)-Q\left(z\right)\hat{M}\left(z\right))\end{array}---\left(16\right)$

Build the Youla Parameterization Controller based on observer

$u\left(z\right)=F\hat{x}\left(z\right)-Q\left(z\right)r\left(z\right)+y_d\left(z\right)$

$\hat{x}(k+1)=A\hat{x}\left(k\right)+\mathrm{Bu}\left(k\right)+\mathrm{Lr}\left(k\right)$

(17)

$\hat{y}\left(k\right)=C\hat{x}\left(k\right)+\mathrm{Du}\left(k\right)+w\left(k\right)$

$r\left(k\right)=y\left(k\right)-\hat{y}\left(k\right)$

Wherein u (z) is gating matrix, y _dz course value that () sets for boats and ships, for the estimated value of system state, for the estimated value that system exports, A, B, C, D is the system matrix built, r (k) is residual matrix, and w (k) is sea wave disturbance, and Q (z) is Youla parameterization matrix, F is state gain, it needs to be configured according to controllable pair (A, B), and fundamental purpose is exactly to ensure system stability.

On-line operation Youla Parameterization Controller, collects inputoutput data u (z, ρ _i), y (z, ρ _i) and residual signals r (z) design weight matrix W _u, W _y, then calculate inputoutput data to ρ _igrad the gradient of input/output signal to parameter ρ is expressed as

$\▿u(z,\mathrm{\ρ})=-M\left(z\right)\▿Q(z,\mathrm{\ρ})r\left(z\right)$

(18)

$\▿y(z,\mathrm{\ρ})=-N\left(z\right)\▿Q(z,\mathrm{\ρ})r\left(z\right)$

Calculate quadratic performance index functional gradient value

${\▿J}_N\left({\mathrm{\ρ}}_i\right)=\frac{1}{N}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}(y^T(k,\mathrm{\ρ})\▿y(k,\mathrm{\ρ})+u^T(k,\mathrm{\ρ})\▿u(k,\mathrm{\ρ}))$

According to gradient descent method, calculate the parameter value ρ of i+1 time _i+1

${\mathrm{\ρ}}_{i+1}={\mathrm{\ρ}}_i-{\mathrm{\γ}}_i{R}_i^{-1}\▿J\left({\mathrm{\ρ}}_i\right)---\left(19\right)$

Wherein $R_i=\frac{1}{N}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}[{\▿y}^T(k,{\mathrm{\ρ}}_i)\▿y(k,{\mathrm{\ρ}}_i)+{\▿u}^T(k,{\mathrm{\ρ}}_i)\▿u(k,{\mathrm{\ρ}}_i)]---\left(20\right)$

If target function J _n(ρ _i) do not reach designing requirement or do not restrain, so, i=i+1, again by input/output signal, the convergence of test rating function, otherwise iteration terminates.

$J_N\left(\mathrm{\ρ}\right)=\frac{1}{2N}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}({\tilde{y}}^T(k,\mathrm{\ρ})W_y\tilde{y}(k,\mathrm{\ρ})+u^T(k,\mathrm{\ρ})W_{u}u(k,\mathrm{\ρ}))---\left(21\right)$

Wherein $\tilde{y}(k,\mathrm{\ρ})=y(k,\mathrm{\ρ})-y_d\left(k\right)$

When target function convergence after, illustrate that Youla Parameterization Controller parameter is optimized for optimum, this Time Controller is in Optimal Control state, input signal through controller process, output order rudder angle.Ship steering engine mathematical model.Usual characteristic of steering gear model is described as:

$T_C\stackrel{\·}{\mathrm{\δ}}=K_C({\mathrm{\δ}}_C-\mathrm{\δ})---\left(22\right)$

Wherein, δ _cfor the order rudder angle assigned; T _cfor the steering wheel time constant of regulation, get 2.5s here; K _cfor servos control gain, experience value is 1; The restrictive condition of ordered rudder angle and rudder speed is | δ | and≤30 °, | δ |≤20 °/s.

Ordered rudder angle generates actual rudder angle after steering wheel process, controls ship motion.In order to the interference that can effectively kill the sea, the present invention considers ship horizontal motion nonlinear mathematical model emphatically:

$m(\stackrel{\·}{u}-\mathrm{vr}-x_G{r}^2+z_G\mathrm{pr})=X$

$m(\stackrel{\·}{v}+\mathrm{ur}+x_G\stackrel{\·}{r}-z_G\stackrel{\·}{p})=Y---\left(23\right)$

$I_z\stackrel{\·}{r}+{\mathrm{mx}}_G(\stackrel{\·}{v}+\mathrm{ur})=N$

Wherein u is surging speed, for surging acceleration, v is swaying speed, for swaying acceleration, r is angular velocity in yaw, for angle of yaw acceleration, p is angular velocity in roll, for roll angle acceleration, for roll angle, x _gfor boats and ships barycenter is to x-axis distance, z _gfor boats and ships barycenter is to z-axis distance, I _zfor hull is to the moment of inertia of y-z plane, I _xfor hull is to the moment of inertia of x-y plane, for the water discharge of boats and ships, g is acceleration of gravity, and ρ is the density of water, the distance of center of gravity G to metancenter M, recover the arm of force.And X, Y, N and K represent hydrodynamic(al) force and moment respectively, they are nonlinear functions of ship motion variable and controlled quentity controlled variable.

Linearization process is carried out to ship nonlinear model

Ship motion equation is nonlinear, but directly to carry out Controller gain variations to nonlinear model be very difficult, designs linear controller, then apply it in the middle of nonlinear model after therefore needing to carry out linearization to model.Due to surging and swaying, rolling and head shake between coupling very weak, surging speed u=0 can be made.5 states are only had like this in linear model and ignore the high-order term coefficient of more than its single order, nonlinear model is through Taylor series expansion, and the linear model obtained after simplification is:

$\stackrel{\·}{x}=\mathrm{Ax}+\mathrm{Bu}=E^{-1}\mathrm{Fx}+E^{-1}\mathrm{G\δ}---\left(24\right)$

$E=\left[\begin{array}{ccccc}m-Y_{\stackrel{\·}{v}}& {\mathrm{mx}}_G-Y_{\stackrel{\·}{r}}& -{\mathrm{mz}}_G-Y_p& 0& 0\\ {\mathrm{mx}}_G-N_{\stackrel{\·}{v}}& I_{\mathrm{zz}}-N_{\stackrel{\·}{r}}& -N_{\stackrel{\·}{p}}& 0& 0\\ -{\mathrm{mz}}_G-K_{\stackrel{\·}{v}}& -K_{\stackrel{\·}{r}}& I_{\mathrm{xx}}-K_{\stackrel{\·}{p}}& 0& 0\\ 0& 0& 0& 1& 0\\ 0& 0& 0& 0& 1\end{array}\right]$

Wherein, E is inertial force system matrix number, and F is viscous force matrix of coefficients, and G is rudder force coefficient matrix.A=E ^-1f is coefficient of combination matrix, B=E ^-1g is system input matrix, and u=δ is the input quantity of boats and ships, i.e. rudder angle.

The yawing that ship motion produces, roll angle is after compass is measured, be input in the Residual Generation device based on observer class, output signal with Time Controller is also input in Residual Generation device, Residual Generation device is to after these signal transacting, generate bow, roll behavior estimated signal and residual signals, and they are input in Youla Parameterization Controller, now whole Control loop system is formed.

Claims (1)

1., based on a control method for the suppression ship course disturbance of data-driven, it is characterized in that, comprise the steps:

(1) the Residual Generation device module input based on state observer class receives actual ship's navigation bow, roll angle signal, control signal u, exports roll angle residual signals with yaw angle residual signals r _ψ:

State observer is $\hat{x}(k+1)=A_x\hat{x}\left(k\right)+B_{x}u\left(k\right)+L_{x}y\left(k\right),$ Residual Generation device is

wherein for the state estimation matrix of state observer, the residual matrix that r (k) is state observer, A _x, B _x, L _x, G, C _x, D _xfor the matrix of coefficients of state observer, u (k) is ordered rudder angle, the yaw angle that y (k) is boats and ships real navigation and roll angle;

(2) two kinds of residual signals and course expectation value are through the process of Youla Parameterization Controller:

Youla Parameterization Controller is wherein u (z) is gating matrix, y _dz course value that () sets for boats and ships, r (z) is residual matrix, and Q (z) is Youla parameterization matrix, and F is state gain matrix;

(3) Youla Parameterization Controller output terminal connects the input end of Rudder Servo System, the control signal that the output terminal of Rudder Servo System exports is input in ship horizontal motion reference model module, the actual ship bow that ship horizontal motion reference model module exports, rolling signal and controller output signal are linked in the Residual Generation device module based on state observer class simultaneously.