Ϣ
Information and Control
1999ꡡ285ڡVol.28No.51999



ĵԪʵʱƷϵͳʵ
±׼

ժҪֽ̽綯ԴųʶĻϣ˴ĵԪʵʱƽ綯ķʵ÷ԷϵͳĿнϸߵĿ³. 
ؼʣԴųʶ INUC㷨 ѧϰ ³
ͼţTP13ױʶ룺B

STUDY ON METHOD OF NONLINEAR SYSTEM PRACTICAL-TIME
CONTROL BY INTELLIGENT NEURAL UNIT
HOU Yuan-bin1 YI Ji-Kai2
(1. Automation of Dept., Xi'an Mining Institute; Xi'an, 710054; 2. Beijing Polytechnic Univesity 100022)
Abstract: The identification of ac motor nonlinear magnetic f ield has been studied, then the method of ac motor practical-time control by in telligent neural controller is proposed in this paper. It has been verified by e xperiment that, the method have higher anti-disturbance ability and strong characteristics for nonlinear system control. 
Key words: nonlinear magnetic identification, algorithm of intelligence single neural control, self-study, strong characteristic

1
ںܶ๤ҵжõ綯϶PIDһֳĿƲԣṹҾһ³ԣڽ綯ϵͳУϵͳĲױ䡢ģͲȷи߶ȵķԣPIDĲߵ˵ϵͳԱӦĿ㷨Ͽϵͳܣ㹤󣬶ڿӦĴϵͳʵ൱ѣ
ȶо[1]ǰֻҪڵļǷǶʽģͿԱƽϵ⺯[4]ָǰԱƽԵĺӣ㹤߱ʶʱֻٶҪ󲻸߷ϵͳ߱ʶо˿ƽĹ㷺ȤŬ̽ڴϵͳеӦãǰɼѾɹԤ⡢ϼϵȿƣ㹤󣬼ٶ޷϶ڷԵٶҪߵʵʱϵͳĿǰֵڷʵʱƵ ڷоϣԽ綯ϵͳķص㣬һִ ĵԪƽ綯ķ(A Method of Intelligence Neural Unit Contro l AC MotorΪINUC)ֿϵͳѧϰ֯ ݽ綯̬INUCĸȨֵߵԽ綯Ķ̬Ȩֵѧϰ̬ƣͻ˳PID̬ܶ仯ֻʵԿƵľԣʵ֤÷ԷϵͳĿнϸߵĿ³ԣ
2INUC㷨
ԪǸ߶ȷԲдϵͳһ񾭽ڵ㣬ʵһ ĶԴԪѧģͿдΪ:
(1)
xi(i=1,2,,n)ΪԪյϢwi(i=1,2,,n)ΪӦȨأΪֵ ұǸֵyΪԪf(.)ǷԵתǵ綯ķ ԣѡöԳԵsigmoid
(2)
ԪͨıͻȨؽѧϰ֯ʹ֪ʶӦı仯ԪõӦԺ³ԣԪѧϰʾΪ
wi(k+1)=wi(k)+[wi(k)-wi(k-1)](3)
ʽЦΪѧϰӣݱضĶ̬0-1֮ȡֵ
2.1 INUCĽṹ
INUCĽṹͼ1ʾͼе߿ΪINUC

ͼ1INUCϵͳṹͼ

2.2 Ԫܻ
ͼеĹжһǸϵͳ̬Զ̬ؿƸȨֵ 11⣬ǽĦ1ʵʵĦ2Ĳֵe״̬任ϵͳԪ״̬ ΪΪ
(4)
ӶɽԪΪ
(5)
(6)
wi(k+1)=wi(k)+i[wi(k)-wi(k-1)](7)
ɹϵͳ̬Զ̬ؿƸȨֵǵʽ(4)(7)ϵͳ eϴʱæ2ͬʱæ3С(ʽ(4)ɼ൱΢ϵСϵ)Ӷ֤ϵͳϵͳeСʱæ2ͦ1Ϊ㣬ͬʱ3Сʹϵͳƽȹɣɹ̬ƸȨֵѧϰ(ԪI NUC)INUCƣÿֱӺͽ綯ĽУǿѧ ϰӦͿ
2.3 ųʶ
2.3.1 MBP㷨Ĵųʶ
MBP㷨[2,5,6]ָĽBP㷨ųʶMBP㷨ȸʵĵ߱ʶȻ󽫱ʶ߱ʶӦãMBPǿĿԺڵļηֲصBPһָĽ㷨ҪĸĽһһֿɱֲСĳʼ㷨Ȩֵֵ󸳳ֵ翹(ȡݴԼԶԵģ𣬵໥һԼ֮)ƵóСŶȨֵ仯ʱĩһȨֵӰѵиѵ̬ĩһȨֵֵ
ݸӦ綯ۼֿƵķ[2]/任ԭ첽綯ĴųģΪ
(8)
ھ̬µ߱ʶMBP㷨BP㷨Ը߷Եĸپȸٶȿ죮ʵһ˵ߵǰںߣ ųʶǾ3/2 任̬ĺ͵Ķ̬ٶȣʶõų
2.3.2 
ʶõĴųMئ2عϵõ̬еתأУǵ̬еתضINUCӰ죬ٿǵƵƵѹҪĵѹȼۺõźuk
ֿƷʽص㣺
.Ԫڶ̬оѧϰԵȳPID иǿӦԣ
.ϵͳйж̬ƸȨֵѧϰʹԪINUCINUCPIDиǿĿ
.ʶʹϵͳԽ綯еķԼر仯Ӱǿʹϵͳиߵ³ԣ
3ʵ
250W첽綯(Ie=0.66AѹUe=380V) вԣԾ̬Ϊֱöÿr1=34.3837( )ýöÿL1=0.276(H)ͣ÷ֹ GD2=2.416(N.M),תJ=GD2/4g=0.06163(N.M/kg)MȡL1ʮ ֮һ
Ϊ֤INUC㷨Чԣȶϵͳ˷森ڶ̬Уڽ綯iA,iB,r2,L2Ƕ̬仯ģʹõģͲȷˣͼ2 (a)αźżϵͳУԴģϵͳĶ̬Եı仯ͼ2(b)(c) PIDŶ½綯ƵķINUCŶ±뽻 綯Ƶķ

(a) αźš (b) PIDƵķ (c) INUC ԱϵͳƵķ
ͼ2 㷨Խ綯Ƶķ

ڶϵͳʱӿõHY-6070A/DD/Aת ƵһԶȵĹ·ƵձʿFRENIC5000G9Sĸܵ Ƶͼ3ΪϵͳʵʵĲ(66M486,TC)ͼ3(a)(b) ڸٶn=1450r/mPIDINUC㷨ƽ綯ʵƬ (c) ڱ为ʱPIDԽ綯ƽ(d) ڱ为ʱINUC㷨Խ綯 ƽ

(a) PIDԽ綯ƽ (b) INUCԽ綯ƽ

(c) ڱ为ʱPIDԽ綯ƽ (d) ڱ为ʱINUC㷨Խ綯ƽ
ͼ3 㷨Խ綯ƽʵƬ

۷INUC㷨ϵͳԶϵͳĲӰ죬ϵͳʵнһ֤INUC㷨PIDԷԵĽ첽иõĿԺ³ԣʵĵƬɿ⸺ʱINUCƵĵʱts3.6룬(ʵʱʾһСΪ3룮֮ǴӰʼбPIDʱ䳤Լ1룬ٶߣ⽫Ϊ)PIDĵʱts4.8룬ڱ为ʱƽȣǰ߱Ⱥ߾иõ³ԣĿǰֵƶڷʵʱƵоڷоϵľ森Ϊ綯ܻṩһ;
עͣĿнίѧչ
߼飺±Ů1955ʿڣαԺƹѧƴͷˣооΪԿƣƣܿƵȣ
׼ǣУ1938ڣΣоΪģƣ㣬ܿƵȣ
ߵλ±ҵѧԺԶϵ710054; ׼ǣҵѧ100022
ο
1Chen T P, & H Chen. Approximation to Continuous Functional by Neur al Networks with Application to Dynamic Systems, IEEE Tran.on Neural Networks, 1994, 5
2±. һԶ̬ϵͳģо. ͨѧʿѧλ, A 397339,97.6
3±, . һ߹ԪԶȵķ. Ӧо1996, (2)
4. ԴݺӦ.ӿƼѧ磬1992
5M D Emmerson, R I Dammper. Determing and Improving the Fault Torlance of Mult ilaer Perception in a PatternRecognition Application, IEEE Trans. Neural Networ ks,1993, 4(5)
6L Holmstorm, P Koisition. Using Additive Noise in Back-Propagation Traing,IEE E Trans. Neural Networks, 1992, 3(1)
ոڣ1998-03-16
