Զѧ
ACTA AUTOMATICA SINICA
1998 24 1  Vol.24  No.1 1998



ģģ߱ʶϵͳ
½ڡϯԣ
ժҪģʵַϵͳ߱ʶ. ȸ˼򻯵ģ﷽ӦӦģ ڴ˻ϸģģͲ߱ʶ㷨. Էģͽ߱ʶ ֤˱ģģͼ߱ʶ㷨. 
ؼʡģϵͳ ϵͳģ߱ʶ㷨. 
ON-LINE IDENTIFICATION OF NONLINEAR
SYSTEMS USING FUZZY MODEL
CHEN JIANQINXI YUGENG
(Institute of Automation, Shanghai Jiaotong University, Shanghai 200030)
AbstractNonlinear system modeling using fuzzy system is discussed in this paper. A simple fuzzy rule representation method and the corresponding adaptive fuzzy inference are proposed. Based on this modified fuzzy system, the paper presents an on-line identification algorithm with which the on-line parameter estimation of nonlinear system is realized. A nonlinear model, as an example, is identified to verify the fuzzy system and identification algorithms.
Key wordsFuzzy systems, nonlinear system modeling, on-line identification, clustering algorithms.
1
ģϵͳ֪ʶʽ  ϵͳģģܵ˹㷺. ģģʵģһѧʽǶģͳһ. ھֲԺģģͣ1ݣ ǰģ ۲ԺԾֲԻΪͨģʵϵͳȫֵķ. ģʹ˾ֲԺܿ˷ģϵͳĸά⣬ѳǹ㷺ʹõģģ. ȻֻھֲԺģģͽģ㷨һЩ⣬磬㷨ܸӣ Բû߱ʶ㷨. 
һģϵͳ֪ʶ£þȷݱķϵͳ Ľеģģͣ ڴ˻ϣ߱ʶ㷨. 
2ģģͺӦģ
ģģʵַϵͳģĶǾȷݣ ԣģͿԼ򻯳ģ
if x is vi,ri,then yi=Tiz, i=1,2,,q(1)
ģ
(2)
Сx=[x1,x2,,xm]Tϵͳ룬z=[1,x1,x2,,xm]Ti=[ai0,ai1,,aim]TǵiӦĽ۲viǵiӦֲģriӦİ뾶sixڵiĳ̶ȣyiǹӦqʾ. 
ģ(1)ϵͳʵ϶ϵͳֲԻȻֲñʾ. xҵһ򼸸ֲʹôĲ͸ռ. ֮ĲͲܰȫռ. һ˵ÿ뷶ΧΪȷôʹ䲢ռ. ÿӦĲ{vi,ri}ͨʶõģôѵܲռ䣬ռһֺֻܻܶ࣬õ֮СռΪӦģµĳĳʱͨķʽ(2). µx(t)κιӦʱ²Ӧģ. 
 adapi(0)=ri(i=1,2,,q),k=1;
 adapi(k)=adapi(k-1)+ri(i=1,2,,q);
if (3)
 if
if (4)
ǣԦriΪÿİ뾶ֱĳһĳЩµx(t)x-vi(1+Ʀ)ri. ÿĽ۲Ժ ʵÿİ뾶ܱ֤ÿĽһ׼ȷ. 
3ģģ߱ʶ
ʽ(1)ģ߱ʶĿҪݲϵõʵʱعƹ{vi,ri}Լi. һҪݵĽṹȷ. ΪˣĲþ㷨ռӦ֣㷨£2ݣ
1) ѡq(Ŀ) ʼĵvi(i=1,2,,q);
2) һx(t)ܶȷĵvc
(5)
У pjʾĵvj̶ܼȣ njΪvjѡΪĴ. 
3) vc
(6)
ʹpjx-vjбȽϿʹpj뾺ʹоᣬӶһ㾺ѧϰе. 
ri(i=1,2,q)ĴСһڹص̶. ģϵͳУ֮໥صǺҪģ˹ģ. Ϊʵ߹Ƹİ뾶ȸϣ֮صȦң Ȼʽȷ뾶
(7)
Ϊ߼ٶȣ ĳһʱֻ̿ȷ
(8)
ڹ۲ԺԵʱx(t)ĳһĳЩʱӦĽ۲ֲi(i=1,2,,q)õС˹£
if
then
(9)
ifx(t)-viri,
then
(10)
ڹƦiʱÿΦiǷҪѵǷĳһĳЩ. ѵ{x(t),y(t)}ڶ뻮ôͬʱi. 
 Բ߹㷨ʵʱģĲvi,ri,i.
㷨1.
1. ȷqصȦ(1ң2). 
2.ʼĵviiʹpi=I,ni=1,i=1,2,,q. һϴIǵλ. 
3.ѵ{x(t),y(t)}. 
4.Ȱʽ(5)ȷĵvcȻʽ(6)vc. 
5.ʽ(7)ri(i=1,2,,q)߰ʽ(8)ֻrc. 
6.ʽ(9)ʽ(10)i. 
7.ת3.
߱ʶ㷨УµɵݶԲƶṩͬϢ. ĳЩѵܵȾʱʹµݲϸ£ɵݶԲƲĲӰȻܱ. ʱԲüӵĵС㷨˷һ. 
4
Ĳģͱƽָʶľȷ̶ȣy(t)ģ͵ʵ y*(t)ǹ. 
.ѡMackey-Glassģͣ3
(11)
бʶ. ѡȡģģx(t+1)=f(x(t),x(t-6),x(t-12),x(t-18))ƽMackey-Glassʱ. ʽ(11)3000ݣ502500ڲƣ25013000ڼģ;. ѡȡ=0.5,q=10. ʵɵPI=0.00057 ͼ1ʾ. ǷǳģʵСͼѺѷֱ棬ùĿҪף3ٵö. 


ͼ1=0.5q=10ʱʵ(ʵ)()ȽϽ. 
5
Էϵͳʶ⣬ĸ˼򻯵ģģͲӦģĸͷ. þ㷨ѧϰģģԤȹ涨Ĺ֮صȷÿ뾶 RLS㷨ʵֹ۲ֲ߹. ĸ߱ʶ㷨ƽȸߣԺãڹ⣬ùĿ٣ʶ㷨еĿѡԱʶѡ. 
1)ȻѧйʿĿ. 
ߵλϺͨѧԶоϺ200030
ο
[1]Takagi T, Sugeno M. Fuzzy identification of systems and its applications to modeling and control. IEEE Trans. Syst., Man, and Cybern., 1985, SMC-15(1):116-132.
[2]Ahalt S C, Krishnamurty A K, Chen P, Melton D E. Competitive learning algorithms for vector quantization. Neural Networks,1990,3(2):277-291.
[3]Wang L X, Mendel J M. Generating fuzzy rules by learning from examples.IEEE Trans. Syst., Man, and Cybern., 1992,SMC-22(6):1414-1427.
ոڡ1996-04-04
