• 検索結果がありません。

Optimization and Design of Maglev System PID Controller Based on Particle Swarm Optimization Algorithm

N/A
N/A
Protected

Academic year: 2021

シェア "Optimization and Design of Maglev System PID Controller Based on Particle Swarm Optimization Algorithm"

Copied!
8
0
0

読み込み中.... (全文を見る)

全文

(1)

书 书 书

50

 

2015

       

西

 

 

 

 

 

 

 

JOURNAL OF SOUTHWEST JIAOTONG UNIVERSITY        Vol. 50  No. 1 Feb. 2015

稿

20140427

:四

2012GZ0103

);教

NCET110712

:宋

1979 -

),女

,副

,博

,研

,电

18080043932

Email

smissmwh@ 163. com

:宋

,陈

PSO

PID

西

2015

50

):

3643.    

02582724

2015

01003608    DOI

10. 3969 / j. issn. 02582724. 2015. 01. 006

PSO

PID

 

1.

西

气工程学院

,四川成都

610031

2.

西南民族大学计算机科学与技术学院

,四川成都

610041

3.

西

,四

川成

610031

 

:为

线

,利

线

线

,设

、线

线

,选

,得

0. 5

,电

0. 49

,并

“好

”,且

“好

”的

0. 561 9.

,优

:磁

;微

;比

;粒

;模

TP273   

Optimization and Design of Maglev System PID Controller Based on

Particle Swarm Optimization Algorithm

SONG Rongrong1

  CHEN Zili3

1. School of Electrical Engineering

Southwest Jiaotong University

Chengdu 610031

China

2. School of Computer Science and Technology

Southwest University for Nationality

Chengdu 610041

China

3. School of Mathematics

Southwest Jiaotong University

Chengdu 610031

China

Abstract

In order to improve the nonlinear and unstable characteristics of maglev systems

two nonlinear subsystems with different structures were transformed into linear ones with the same structure by differential geometry method

and then a PID controller that is based on the standard particle swarm optimization

PSO

was built. For the PSO algorithm

the fixed inertia weight

FIW

),

linear descend inertia weight

LIW

),

and linear differential descend inertia weight

LDW

were comparatively studied through simulation

among them the FIW was found more suitable and hence selected for electromagnet 1 and electromagnet 2. After optimization of controller parameters

the value of the FIW parameter C for electromagnets 1 and 2 was set to 0. 5 and 0. 49

respectively. In addition

a fuzzy comprehensive evaluation model was built to evaluate the antiinterference performance of the electromagnets 1 and 2

revealing that the two electromagnets both have a good antiinterference ability

and the membership degree of

good

is 0. 561 9. Experimental results show that the optimized maglev system has a fairly good robustness. Key words

maglev system

differential geometry

proportionalintegraldifferential controller

particle swarm optimization algorithm

fuzzy comprehensive evaluation    

性 能

,运

500 km / h

,显

(2)

,等

:基

PSO

PID

,其

,悬

,研

23

. PID

proportionalintegraldifferential

)控

. PID

:比

KP

、积

KI

KD.

PID

,国

PID

510

]提

PID

,该

线

,自

PID

,但

,若

,则

,而文中并没有进行参数优化选取

10

]为

减小磁浮列车气隙控制中非线性的影

,将

,并

线

,提

,该

、停

,以

,但

10

,常

线

,忽

,得

线

PID

,基

线

,在

,当

,控

,甚

,本

、比

于控制器设计

优化了基于标准

PSO

particle swarm optimization

)算法的

PID

控制器

,通过阶跃

,选

;同

,利

,对

1 

1. 1 

   

受力分析如图

所示

,取向上为正方

,电

m¨c1 = F11 - F21- mg - f12- k1c1

m¨c2 = F22- F12 - mg + f12 - k1c2

c1

c2

线

F11

F21

线

F12

F22

线

f12

k1

. 































              



      

1 

Fig. 1  Force analysis of maglev system

11

F11= u1/

c1 + b

],

F21= u2/

l + c2 + b

],

F22= u2/

- c2 + b

],

F12= u1/

l + c2 + b

],

u1

u2

线

、线

线

线

f12

11

f12 = c /

c12+ d

c12= l + c2 - c1

c12

. 1. 2 

线

   

F12

F21

f12

,选

x1 = c1

  x2 = c1

  x3 = c2

  x4 = c2

状态空间为(

)、

)、

)参见文

11

= F

x + G

  y = H

x.

(式

))中

y1

12

y1 = Lfhj

2 i = 1

Lgihj

))

ui = x2

, (

10

Lgihj

≡0. 7 3

(3)

西

 

 

 

 

 

 

 

50

r1

c1

依赖系统输入的最小整数

,则

r1 = 2.

r2 = 2

c2

12

]知

,系

(式

))的

r1

r2

),且

r1 + r2 = 4.

,由

线

12

,系

(式

))可

线

= A- 1

)[

- B

+ V

],其

A- 1

= ma

x1 + b

4 0 0 ma

- x3 + b

[

]

= - g - c2x2 - g - c2x

[

]

(式

))经

线

,状

= 0 1 0 0 0 0 0 0 0 0 0 1           0 0 0 0 x1 x2 x3 x             4 + 0 0 1 0 0 0           0 1 V1 V

[ ]

y = H

x.

11

1. 3 

线

   

线

线

,传

G1 = G2 = 1 / s 2

12

12

step

13

,由

,需

.           







 



      





2 

线

Fig. 2  Unit step response of linear system

2 

于标准粒子群算法的

PID

控制

   

法(

particle swarm optimization

PSO

)是

,通

,行

,每

3.

PSO

PID

14

1 

、最

kmax

、粒

vmin

vmax

],初

pi

vi

,找

,并

gmax

pmax.

2 

,记

vi

vi1

),

vi2

),

vi3

))

13

pi

pi1

),

pi2

),

pi3

))

T .

14

J =

∞ 0

z1 e

+ z2u 2

+   z4 Δe

) )

dk + z3ku

15

)为

)为

ku

z1 ~ z4

14

],

z1 = 0. 999

z2 = 0. 001

z3 = 2

z4 = 100.

,则

;如

,则

3 

k + 1

vi

k + 1

= w

vi

+ S1φ1

pibest

- xi

))

+   S2φ2

pbest

- xi

)),

16

xid

k + 1

= xid

+ vid

),

  1≤i≤M

,(

17

S1

S2

φ1

φ2

)区

)为

vi > vmax

,则

vi

vmax

vi < vmin

,则

vi

vmin.

) 固定惯性权重(

fixed inertia weight

FIW

14

= C = 0. 68.

18

)线

linear descend inertia weight

LIW

14

= wstart -

[(

wstart- wend

/ kmax

, (

19

wstart

LIW

wend

LIW

) 线性微分递减惯性权重(

linear differential descend inertia weight

LDW

14

= wstart -

[(

wstart- wend

/ k 2 max

k 2 .

20

4 

8 3

(4)

,等

:基

PSO

PID

kmax

,输

;否

,转

2.

3 

   

PSO

,有

,收

,但

;当

,有

,得

,但

,因

,调

3.

1 

Tab. 1  Control performances

重稳

/ mm

/ %

/ s

/ s

重稳

/ mm

/ %

/ s

/ s FIW 5 87. 85 0. 638 9. 622 FIW 5 22. 60 2. 841 18. 915 LIW 5 90. 30 0. 637 9. 765 LIW 5 12. 01 3. 576 34. 100 LDW 5 90. 83 0. 637 13. 606 LDW 5 11. 11 3. 261 33. 494            























  



           























 



FIW            























 



           























 



LIW            























  



           























  



LDW

3 

线

Fig. 3  Step response of linear system 9 3

(5)

西

 

 

 

 

 

 

 

50

   

t = 0. 001 s

PSO

:粒

30

,惯

0. 9

,惯

0. 4

,最

100

S1 = S2 = 2

z1 = 0. 999

  z2 = 0. 001

z3 = 2

  z4 = 100

14

,针

FIW

,上

,惯

FIW

;针

FIW

LDW

FIW

LDW

都较适合

,选择惯性权重

FIW. FIW

,观

(表

,针

,当

C = 0. 50

,超

,因

0. 50

,当

C = 0. 49

,超

,因

0. 49

2 

Tab. 2  Control performance of different C

1 C

/ mm

/ %

/ s

/ s

2 C

/ mm

/ %

/ s

/ s 0. 40 5 86. 46 0. 637 8. 620 0. 40 - 20 35. 75 1. 518 13. 383 0. 45 5 87. 08 0. 637 8. 787 0. 45 - 20 32. 80 1. 592 10. 392 0. 50 5 83. 73 0. 637 9. 463 0. 49 - 20 1. 98 4. 212 7. 043 0. 60 5 88. 78 0. 637 9. 207 0. 50 - 20 11. 02 3. 260 33. 635 0. 70 5 92. 01 0. 637 13. 732 0. 60 - 20 15. 95 3. 375 27. 258 0. 70 - 20 35. 91 1. 498 12. 943    

使

PID

t = 0. 001 s

,自

PID

:种

30

,交

0. 900

,变

0. 033.

PID

仿

,与

,优

PSO

PID

,比

24%

64%

,上

PID

3  PID

Tab. 3  Optimized parameters of the PID controller

数积

数微

数积

数微

6. 337 9 0. 962 5 0. 015 7 5. 767 4 0. 976 5 0. 976 5

PSO

6. 245 9 1. 545 7 3. 134 4 2. 671 4 5. 875 6 0. 023 9

4 

PID

Tab. 4  Control performance of different PID controllers

/ %

/ s

/ s

/ %

/ s

/ s

110. 000 0. 750 10. 654 5. 500 6. 414 12. 353

PSO

83. 730 0. 637 9. 463 1. 980 4. 212 7. 043

4 

4. 1 

   

、低

、高

线

5 

线

Tab. 5  Evaluation indexes of the linear system

U1

U11

U12

U2

U21

U22

U3

U31

U32

U4

U41

U42 0 4

(6)

,等

:基

PSO

PID

4. 2 

PSO

PID

   

PSO

PID

,设

,评

6 

线

Tab. 6  Controller parameters of the linear system

铁 比

数 积

数 微

1 6. 245 9 1. 545 7 3. 134 4 2 2. 671 4 5. 875 6 0. 023 9 4. 3 

   

、准

,两

15

,建

,见

4.

,构

线

P = 1 3 5 7 1 / 3 1 3 5 1 / 3 1 / 3 1 3 1 / 7 1 / 5 1 /           3 1 .

P′ = 1 3 1 /

[

3 1

]

7 

线

Tab. 7  Evaluation index values of the linear system

2 U1 - 0. 0011 0. 201 5 0. 007 4 0. 309 4 U2 - 0. 001 4 0. 060 4 - 0. 021 9 0. 201 4 U3 - 4. 100 7×10- 6 0. 019 0 - 0. 012 3 0. 483 3 U4 0. 002 5 0. 002 1 0. 037 7 0. 308 5   





















 







 







 







 







 











4 

Fig. 4  Hierarchical structures of evaluation indexes    

,计

的最大特征值相应的特征向量

P = 1 3 5 7 1 / 3 1 3 5 1 / 3 1 / 3 1 3 1 / 7 1 / 5 1 /           3 1 → 

  105 / 120 45 / 68 15 / 28 7 / 16 35 / 120 15 / 68 9 / 28 5 / 16 21 / 120 5 / 68 3 / 28 3 / 16 15 / 120 3 / 68 1 / 28 1 /           16

  

2. 510 0 1. 146 2 0. 543 2 0.           267 3 → 



W = 0. 561 9 0. 256 6 0. 121 6 0.           059 8

λmax =

4 i = 1

PW

i 4Wi = 4. 123 8

ηλ = 0. 041

ηR = 0. 045 9 < 0. 1. P′

W′

P′ = 1 3 1 /

[

3 1

]

→

3 / 4 3 / 4 1 / 4 1 /

[

]

  

3 / 2 1 /

[ ]

2 →

W′ = 0. 75 0.

[ ]

25

λ′max =

2 i = 1

P′W′

i 2W′i = 2

η′λ= 0

η′R= 0 < 0. 1. 4. 4 

   

)确

“好

”、“中

”和

“差

”的

15

1 4

(7)

西

 

 

 

 

 

 

 

50

β1

ζ

= 1

  ζ≤γ1

γ2 - ζ γ2 - γ1

  γ1 < ζ < γ2

  ζ≥γ2       

β2

ζ

= 0

  ζ≤γ1

ζ - γ1 γ2 - γ1

  γ1 < ζ < γ2

  γ2≤ζ≤γ3

γ4 - ζ γ4 - γ3

  γ3 < ζ < γ4

  ζ≥γ2           

β3

ζ

= 0

  0≤ζ≤γ1

ζ - γ1 γ2 - γ1

  γ1 < ζ < γ2

  ζ≥γ2       

ζ

线

γ1

γ2

γ3

γ4

β1

“好

”;

β2

“中

”;

β3

“差

线

8.

8 

Tab. 8  Reference nodes of membership functions

/ mm

/ mm

γ1= 0. 001

γ2= 0. 100 γ1= 0. 001

γ2= 1. 000

γ1= 0. 001

γ2= 0. 004

γ3= 0. 006

γ4= 0. 100 γ1= 0. 001

γ2= 0. 004

γ3= 0. 006

γ4= 1. 000

γ1= 0. 001

γ2= 1. 000 γ1= 0. 001

γ2= 1. 000

)对

,得

R11= 0. 999 0 0. 033 3 1 × 10- 4 0. 799 3 0. 803 3 0.

[

200 7

]

R12= 0. 996 0 0. 133 3 1 × 10- 4 0. 940 5 0. 945 3 0.

[

059 5

]

R13= 1 0 0 0. 982 0 0. 986 9 0.

[

018 0

]

R14= 0. 984 8 0. 500 0 0. 001 5 0. 998 9 0. 366 7 0.

[

001 5

]

R21= 1 0 0 0. 691 3 0. 694 8 0.

[

308 7

]

R22= 0. 991 9 0. 266 7 8 × 10- 4 0. 799 4 0. 803 4 0.

[

200 6

]

R23 = 0. 988 7 0. 993 7 0. 011 3 0. 517 2 0. 519 8 0.

[

482 8

]

R24 = 0. 963 3 0. 968 1 0. 036 7 0. 692 2 0. 695 7 0.

[

307 8

]

)对

B11 = W′ T  R11

B12

B13

B14

B21

B22

B23

B24.

)对

R1 = 0. 75 0. 25 0. 200 7 0. 75 0. 25 0. 059 5 0. 75 0. 25 0. 018 0 0. 75 0. 25 0.           001 5

R2 = 0. 75 0. 25 0. 200 7 0. 75 0. 25 0. 200 6 0. 75 0. 75 0. 250 0 0. 75 0. 75 0.           250 0

B1 = W T  R1 =  

0. 561 9  0. 250 0  0. 200 0

),

B2 = W T  R2 =  

0. 561 9  0. 256 6  0. 200 7

),

,得

是“好

”的

,“好

”的

0. 561 9.

5 

   

)利

,将

线

线

)通

,得

FIW

)比

2.

C = 0. 50

C = 0. 49.

,提

)模糊综合评判模型评价优化后的电磁

“好

”,且

“好

0. 561 9.

)采

,可

 

祥明

磁浮列车[

上海

:上海科学技术出版

2 4

参照

関連したドキュメント

Suppose the basic data are as shown in Section 4.1, no shifting-berth operation exists and all tugboats do not return to the anchorage base during the planning horizon, use the

The performance of scheduling algorithms for LSDS control is usually estimated using a certain number of standard parameters, like total time or schedule

We present the new multiresolution network flow minimum cut algorithm, which is es- pecially efficient in identification of the maximum a posteriori (MAP) estimates of corrupted

In this paper, taking into account pipelining and optimization, we improve throughput and e ffi ciency of the TRSA method, a parallel architecture solution for RSA security based on

We present the new multiresolution network flow minimum cut algorithm, which is es- pecially efficient in identification of the maximum a posteriori (MAP) estimates of corrupted

This paper deals with the a design of an LPV controller with one scheduling parameter based on a simple nonlinear MR damper model, b design of a free-model controller based on

Karzanov: Minimum 0-extensions of graph metrics, Europ.. Metric relaxation (Karzanov

Based on the stability theory of fractional-order differential equations, Routh-Hurwitz stability condition, and by using linear control, simpler controllers are designed to